Datasets:

bigcode

/

jupyter-parsed

like 4

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import os\nimport glob\nimport zipfile\nimport pathlib\nimport cv2\nimport math\nimport random\nimport shutil\nimport skimage as sk\nimport pandas as pd\nimport numpy as np\nimport tensorflow as tf\nimport matplotlib.pyplot as plt\nimport matplotlib.image as mpimg\n\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.preprocessing import MinMaxScaler\nfrom sklearn.metrics import mean_squared_error\nfrom sklearn.utils import shuffle\n\nfrom keras.models import Sequential\nfrom keras.activations import softmax, relu\nfrom keras.layers import Activation, Dense, Dropout, Flatten, Lambda, Cropping2D, LSTM\nfrom keras.layers import Conv2D, MaxPooling2D, BatchNormalization\nfrom keras.optimizers import Adam\nfrom keras.regularizers import l2\nfrom keras.callbacks import EarlyStopping, ModelCheckpoint", "Using TensorFlow backend.\n" ] ], [ [ "## Ran the new few blocks for my colab configuration, can be ignored.", "_____no_output_____" ] ], [ [ "from google.colab import drive\n\ndrive.mount('/content/gdrive')", "Drive already mounted at /content/gdrive; to attempt to forcibly remount, call drive.mount(\"/content/gdrive\", force_remount=True).\n" ], [ "!wget https://d17h27t6h515a5.cloudfront.net/topher/2016/December/584f6edd_data/data.zip\n", "--2019-04-14 10:46:55-- https://d17h27t6h515a5.cloudfront.net/topher/2016/December/584f6edd_data/data.zip\nResolving d17h27t6h515a5.cloudfront.net (d17h27t6h515a5.cloudfront.net)... 52.85.159.80, 52.85.159.38, 52.85.159.165, ...\nConnecting to d17h27t6h515a5.cloudfront.net (d17h27t6h515a5.cloudfront.net)|52.85.159.80|:443... connected.\nHTTP request sent, awaiting response... 200 OK\nLength: 333137665 (318M) [application/zip]\nSaving to: ‘data.zip’\n\ndata.zip 100%[===================>] 317.70M 42.5MB/s in 7.2s \n\n2019-04-14 10:47:03 (43.9 MB/s) - ‘data.zip’ saved [333137665/333137665]\n\n" ], [ "import shutil\n\nshutil.move(\"/content/data.zip\", \"/content/gdrive/My Drive/udacity-behavioural-cloning/\")", "_____no_output_____" ], [ "os.chdir('/content/gdrive/My Drive/udacity-behavioural-cloning/')", "_____no_output_____" ], [ "with zipfile.ZipFile('data.zip') as f:\n f.extractall()", "_____no_output_____" ], [ "os.chdir('/content/gdrive/My Drive/udacity-behavioural-cloning/data/')", "_____no_output_____" ] ], [ [ "## Training code starts here", "_____no_output_____" ] ], [ [ "df = pd.read_csv('driving_log.csv')", "_____no_output_____" ], [ "# Visualizing original distribution\n\nplt.figure(figsize=(15, 3))\nhist, bins = np.histogram(df.steering.values, bins=50)\nplt.hist(df.steering.values, bins=bins)\nplt.title('Steering Distribution Plot')\nplt.xlabel('Steering')\nplt.ylabel('Count')\nplt.show()", "_____no_output_____" ], [ "# create grayscale image\ndef grayscale(img):\n return cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)\n# normalize image to zero mean\ndef normalize(img):\n mean = np.mean(img)\n std = np.std(img)\n return (img-mean)/std\n# preprocess with grayscale and normalization\ndef preprocess(img):\n return normalize(grayscale(img))\n# augment image, left right flip for now\ndef augment(image, randn):\n return np.flip(image, axis=randn%2).astype(np.uint8)\n# yeo-johnson bias\ndef yeo_johnson_bias(steering):\n if steering >= 0:\n return np.log(steering + 1)\n elif steering < 0:\n return -np.log(-steering + 1)", "_____no_output_____" ], [ "# To separate center, left and right\ndf_center = pd.concat([df.center, df.steering], axis=1).rename(index=str, columns={'center': 'img'})\ndf_left = pd.concat([df.left, df.steering], axis=1).rename(index=str, columns={'left': 'img'})\ndf_right = pd.concat([df.right, df.steering], axis=1).rename(index=str, columns={'right': 'img'})\n\ndf_center.head()", "_____no_output_____" ], [ "# Adjusting the steering value 0 for left and right\n\nfor k, v in df_left.iterrows():\n if v.steering == 0:\n df_left.loc[k, 'steering'] = df_left.loc[k, 'steering'] + random.uniform(0.2, 0.5)\n \nfor k, v in df_right.iterrows():\n if v.steering == 0:\n df_right.loc[k, 'steering'] = df_right.loc[k, 'steering'] + random.uniform(-0.2, -0.5)", "_____no_output_____" ], [ "new_df = pd.concat([df_center, df_left, df_right], axis=0, ignore_index=True, sort=False)", "_____no_output_____" ], [ "new_df.tail()", "_____no_output_____" ], [ "new_df.to_csv('adjusted_log.csv', index=False, encoding='utf-8')", "_____no_output_____" ], [ "df = pd.read_csv('adjusted_log.csv')", "_____no_output_____" ], [ "# Visualizing adjusted distribution\n\nplt.figure(figsize=(15, 3))\nhist, bins = np.histogram(df.steering.values, bins=50)\nplt.hist(df.steering.values, bins=bins)\nplt.title('Steering Distribution Plot')\nplt.xlabel('Steering')\nplt.ylabel('Count')\nplt.show()\n\ndf.plot(figsize=(15, 3))", "_____no_output_____" ], [ "df.shape", "_____no_output_____" ], [ "# Grouping all images and steering together, to do a train test splitting\n\nimages = df.img.tolist()\nsteering = df.steering.tolist()\n\nimg_list = []\nfor img, angle in zip(images, steering):\n row = [img, angle]\n img_list.append(row)\n\ntrain_samples, validation_samples = train_test_split(img_list, test_size=0.2)", "_____no_output_____" ], [ "# Data generator\n\ndef generator(samples, batch_size=32):\n cwd = os.getcwd()\n num_samples = len(samples)\n while True: # Loop forever so the generator never terminates\n shuffle(samples)\n for offset in range(0, num_samples, batch_size):\n batch_samples = samples[offset:offset+batch_size]\n\n images = []\n angles = []\n\n for batch_sample in batch_samples:\n name = os.path.join(cwd, batch_sample[0].strip())\n try:\n # normalizing image\n image = normalize(mpimg.imread(name))\n # reshaping image into its rgb form\n image = np.reshape(image, (image.shape[0], image.shape[1], 3))\n steering = float(batch_sample[1])\n images.append(image)\n angles.append(steering)\n # if image not found, skip the image\n except FileNotFoundError as msg:\n print(msg)\n continue\n\n # trim image to only see section with road|\n X_train = np.array(images)\n y_train = np.array(angles)\n yield shuffle(X_train, y_train)", "_____no_output_____" ], [ "# Set our batch size\nbatch_size = 32\n\n# compile and train the model using the generator function\ntrain_generator = generator(train_samples, batch_size=batch_size)\nvalidation_generator = generator(validation_samples, batch_size=batch_size)", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "### PART 3: TRAINING ###\n# Training Architecture: inspired by NVIDIA architecture #\nINPUT_SHAPE = (160, 320, 3)\nmodel = Sequential()\nmodel.add(Cropping2D(cropping=((70,25), (0, 0)), input_shape=INPUT_SHAPE))\n\nmodel.add(Conv2D(filters=24, kernel_size=5, strides=(2, 2), activation='relu'))\nmodel.add(BatchNormalization())\nmodel.add(Conv2D(filters=36, kernel_size=5, strides=(2, 2), activation='relu'))\n#model.add(BatchNormalization())\nmodel.add(Conv2D(filters=48, kernel_size=5, strides=(2, 2), activation='relu'))\n#model.add(BatchNormalization())\nmodel.add(Conv2D(filters=64, kernel_size=3, strides=(1, 1), activation='relu'))\n#model.add(BatchNormalization())\nmodel.add(Conv2D(filters=64, kernel_size=3, strides=(1, 1), activation='relu'))\n\nmodel.add(Flatten())\nmodel.add(Dense(1164, activation='relu'))\nmodel.add(Dropout(rate=0.5))\nmodel.add(Dense(100, activation='relu'))\nmodel.add(Dropout(rate=0.5))\nmodel.add(Dense(50, activation='relu'))\nmodel.add(Dropout(rate=0.5))\nmodel.add(Dense(10, activation='relu'))\nmodel.add(Dense(1))\nadam = Adam(lr = 0.0001)\nmodel.compile(optimizer= adam, loss='mse', metrics=['accuracy'])\nmodel.summary()\n", "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\nInstructions for updating:\nColocations handled automatically by placer.\nWARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\nInstructions for updating:\nPlease use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n_________________________________________________________________\nLayer (type) Output Shape Param # \n=================================================================\ncropping2d_1 (Cropping2D) (None, 65, 320, 3) 0 \n_________________________________________________________________\nconv2d_1 (Conv2D) (None, 31, 158, 24) 1824 \n_________________________________________________________________\nbatch_normalization_1 (Batch (None, 31, 158, 24) 96 \n_________________________________________________________________\nconv2d_2 (Conv2D) (None, 14, 77, 36) 21636 \n_________________________________________________________________\nconv2d_3 (Conv2D) (None, 5, 37, 48) 43248 \n_________________________________________________________________\nconv2d_4 (Conv2D) (None, 3, 35, 64) 27712 \n_________________________________________________________________\nconv2d_5 (Conv2D) (None, 1, 33, 64) 36928 \n_________________________________________________________________\nflatten_1 (Flatten) (None, 2112) 0 \n_________________________________________________________________\ndense_1 (Dense) (None, 1164) 2459532 \n_________________________________________________________________\ndropout_1 (Dropout) (None, 1164) 0 \n_________________________________________________________________\ndense_2 (Dense) (None, 100) 116500 \n_________________________________________________________________\ndropout_2 (Dropout) (None, 100) 0 \n_________________________________________________________________\ndense_3 (Dense) (None, 50) 5050 \n_________________________________________________________________\ndropout_3 (Dropout) (None, 50) 0 \n_________________________________________________________________\ndense_4 (Dense) (None, 10) 510 \n_________________________________________________________________\ndense_5 (Dense) (None, 1) 11 \n=================================================================\nTotal params: 2,713,047\nTrainable params: 2,712,999\nNon-trainable params: 48\n_________________________________________________________________\n" ], [ "history = model.fit_generator(generator=train_generator, steps_per_epoch=math.ceil(len(train_samples)/ batch_size), \\\n epochs=15, verbose=1, validation_data=validation_generator, \\\n validation_steps=math.ceil(len(validation_samples)/ batch_size), use_multiprocessing=False)\n\nprint('Done Training')", "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\nInstructions for updating:\nUse tf.cast instead.\nEpoch 1/15\n603/603 [==============================] - 197s 326ms/step - loss: 0.0531 - acc: 0.1800 - val_loss: 0.0418 - val_acc: 0.1839\nEpoch 2/15\n603/603 [==============================] - 193s 320ms/step - loss: 0.0435 - acc: 0.1801 - val_loss: 0.0393 - val_acc: 0.1839\nEpoch 3/15\n603/603 [==============================] - 194s 321ms/step - loss: 0.0407 - acc: 0.1801 - val_loss: 0.0380 - val_acc: 0.1839\nEpoch 4/15\n603/603 [==============================] - 194s 321ms/step - loss: 0.0381 - acc: 0.1801 - val_loss: 0.0363 - val_acc: 0.1839\nEpoch 5/15\n603/603 [==============================] - 193s 320ms/step - loss: 0.0370 - acc: 0.1801 - val_loss: 0.0369 - val_acc: 0.1839\nEpoch 6/15\n603/603 [==============================] - 193s 319ms/step - loss: 0.0349 - acc: 0.1801 - val_loss: 0.0355 - val_acc: 0.1839\nEpoch 7/15\n603/603 [==============================] - 193s 320ms/step - loss: 0.0331 - acc: 0.1802 - val_loss: 0.0353 - val_acc: 0.1839\nEpoch 8/15\n603/603 [==============================] - 191s 318ms/step - loss: 0.0306 - acc: 0.1801 - val_loss: 0.0352 - val_acc: 0.1839\nEpoch 9/15\n603/603 [==============================] - 191s 317ms/step - loss: 0.0279 - acc: 0.1802 - val_loss: 0.0352 - val_acc: 0.1839\nEpoch 10/15\n603/603 [==============================] - 191s 317ms/step - loss: 0.0254 - acc: 0.1802 - val_loss: 0.0358 - val_acc: 0.1839\nEpoch 11/15\n603/603 [==============================] - 192s 318ms/step - loss: 0.0228 - acc: 0.1802 - val_loss: 0.0356 - val_acc: 0.1839\nEpoch 12/15\n603/603 [==============================] - 191s 317ms/step - loss: 0.0204 - acc: 0.1803 - val_loss: 0.0357 - val_acc: 0.1839\nEpoch 13/15\n603/603 [==============================] - 190s 315ms/step - loss: 0.0179 - acc: 0.1803 - val_loss: 0.0364 - val_acc: 0.1839\nEpoch 14/15\n603/603 [==============================] - 189s 314ms/step - loss: 0.0158 - acc: 0.1804 - val_loss: 0.0374 - val_acc: 0.1839\nEpoch 15/15\n603/603 [==============================] - 189s 314ms/step - loss: 0.0141 - acc: 0.1804 - val_loss: 0.0374 - val_acc: 0.1839\nDone Training\n" ], [ "###Saving Model and Weights###\nmodel_json = model.to_json()\nwith open(\"model5.json\", \"w\") as json_file:\n json_file.write(model_json)\nmodel.save('model5.h5')\nmodel.save_weights(\"model_weights5.h5\")\nprint(\"Saved model to disk\")", "Saved model to disk\n" ], [ "### print the keys contained in the history object\nprint(history.history.keys())\n\n### plot the training and validation loss for each epoch\nplt.figure(figsize=(15, 3))\nplt.plot(history.history['loss'])\nplt.plot(history.history['val_loss'])\nplt.title('model mean squared error loss')\nplt.ylabel('mean squared error loss')\nplt.xlabel('epoch')\nplt.legend(['training set', 'validation set'], loc='upper right')\nplt.show()", "dict_keys(['val_loss', 'val_acc', 'loss', 'acc'])\n" ], [ "", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code" ]

[ [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# entities-search-engine loading\nSPARQL query to `{\"type\": [values]}`", "_____no_output_____" ] ], [ [ "import sys\nsys.path.append(\"..\")\n\nfrom heritageconnector.config import config\nfrom heritageconnector.utils.sparql import get_sparql_results\nfrom heritageconnector.utils.wikidata import url_to_qid\n\nimport json\nimport time\nfrom tqdm import tqdm\n\nendpoint = config.WIKIDATA_SPARQL_ENDPOINT", "_____no_output_____" ] ], [ [ "## humans sample", "_____no_output_____" ] ], [ [ "limit = 10000\n\nquery = f\"\"\"\nSELECT ?item WHERE {{\n ?item wdt:P31 wd:Q5.\n}} LIMIT {limit}\n\"\"\"\n\nres = get_sparql_results(endpoint, query)\n\ndata = {\n \"humans\": [url_to_qid(x['item']['value']) for x in res['results']['bindings']]\n}\n\nwith open(\"./entities-search-engine/data/humans_sample.json\", 'w') as f:\n json.dump(data, f)", "_____no_output_____" ] ], [ [ "## humans sample: paginated\ngot a 500 timeout error nearly all of the way through. Looked like it was going to take around 1h20m. *Better to do with dump?*", "_____no_output_____" ] ], [ [ "# there are 8,011,382 humans in Wikidata so this should take 161 iterations\ntotal_humans = 8011382\npagesize = 40000\nreslen = pagesize\n\npaged_json = []\ni = 0\n\nstart = time.time()\npbar = tqdm(total=total_humans)\n\nwhile reslen == pagesize:\n query = f\"\"\"\n SELECT ?item WHERE {{\n ?item wdt:P31 wd:Q5.\n }} LIMIT {pagesize} OFFSET {i*pagesize}\n \"\"\"\n\n res = get_sparql_results(endpoint, query)['results']['bindings']\n reslen = len(res)\n \n paged_json.append(\n { \"humans\": [url_to_qid(x['item']['value']) for x in res] }\n )\n \n # print total number so far\n #print(i+1, (i+1)*pagesize)\n i+=1\n pbar.update(pagesize)\n \nend = time.time()\npbar.close()\n\nprint(f\"COMPLETED: {round(end-start, 2)} seconds\")\n \n# with open(\"./entities-search-engine/data/humans_sample.json\", 'w') as f:\n# json.dump(data, f)", " 89%|████████▉ | 7120000/8011382 [58:13<12:59, 1143.37it/s]" ], [ "for idx, item in tqdm(enumerate(paged_json)):\n with open(f\"./entities-search-engine/data/humans/humans_{idx}.json\", 'w') as f:\n json.dump(item, f)", "\n0it [00:00, ?it/s]\u001b[A\n3it [00:00, 24.90it/s]\u001b[A\n5it [00:00, 22.01it/s]\u001b[A\n8it [00:00, 23.76it/s]\u001b[A\n11it [00:00, 24.37it/s]\u001b[A\n15it [00:00, 24.48it/s]\u001b[A\n19it [00:00, 26.02it/s]\u001b[A\n22it [00:00, 26.84it/s]\u001b[A\n25it [00:00, 27.48it/s]\u001b[A\n29it [00:01, 28.67it/s]\u001b[A\n32it [00:01, 27.39it/s]\u001b[A\n35it [00:01, 27.23it/s]\u001b[A\n38it [00:01, 25.93it/s]\u001b[A\n41it [00:01, 25.11it/s]\u001b[A\n44it [00:01, 24.59it/s]\u001b[A\n47it [00:01, 23.96it/s]\u001b[A\n50it [00:01, 22.40it/s]\u001b[A\n53it [00:02, 23.28it/s]\u001b[A\n56it [00:02, 24.64it/s]\u001b[A\n59it [00:02, 24.45it/s]\u001b[A\n62it [00:02, 25.65it/s]\u001b[A\n65it [00:02, 25.37it/s]\u001b[A\n68it [00:02, 25.41it/s]\u001b[A\n71it [00:02, 25.86it/s]\u001b[A\n74it [00:02, 26.11it/s]\u001b[A\n77it [00:03, 26.18it/s]\u001b[A\n80it [00:03, 26.45it/s]\u001b[A\n83it [00:03, 26.71it/s]\u001b[A\n86it [00:03, 26.90it/s]\u001b[A\n89it [00:03, 24.88it/s]\u001b[A\n92it [00:03, 25.85it/s]\u001b[A\n95it [00:03, 25.95it/s]\u001b[A\n98it [00:03, 25.79it/s]\u001b[A\n101it [00:03, 25.92it/s]\u001b[A\n104it [00:04, 26.47it/s]\u001b[A\n108it [00:04, 27.56it/s]\u001b[A\n111it [00:04, 27.39it/s]\u001b[A\n114it [00:04, 27.24it/s]\u001b[A\n117it [00:04, 27.55it/s]\u001b[A\n120it [00:04, 27.45it/s]\u001b[A\n123it [00:04, 27.16it/s]\u001b[A\n126it [00:04, 26.65it/s]\u001b[A\n129it [00:04, 26.50it/s]\u001b[A\n132it [00:05, 26.36it/s]\u001b[A\n135it [00:05, 27.20it/s]\u001b[A\n138it [00:05, 27.28it/s]\u001b[A\n141it [00:05, 27.53it/s]\u001b[A\n144it [00:05, 27.68it/s]\u001b[A\n147it [00:05, 27.42it/s]\u001b[A\n150it [00:05, 27.27it/s]\u001b[A\n153it [00:05, 27.94it/s]\u001b[A\n156it [00:05, 26.64it/s]\u001b[A\n159it [00:06, 26.79it/s]\u001b[A\n162it [00:06, 26.72it/s]\u001b[A\n165it [00:06, 26.58it/s]\u001b[A\n168it [00:06, 26.12it/s]\u001b[A\n172it [00:06, 27.28it/s]\u001b[A\n175it [00:06, 27.70it/s]\u001b[A\n178it [00:06, 26.39it/s]\u001b[A\n" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "from google.colab import drive\ndrive.mount('/content/gdrive')", "Drive already mounted at /content/gdrive; to attempt to forcibly remount, call drive.mount(\"/content/gdrive\", force_remount=True).\n" ], [ "import pandas as pd\nimport numpy as np\nimport csv\n\n#DATA_FOLDER = '/content/gdrive/My Drive/101/results/logreg/'\n\nsubfolders = []\nfor a in range(1,7):\n for b in range(6,0,-1):\n subfolders.append('+1e-0'+str(a)+'_+1e-0'+str(b))\n\nclassifiers = ['logreg', 'mlp', 'better_cnn']\nall_results = []\n\nfor clf in classifiers:\n print(clf)\n DATA_FOLDER = '/content/gdrive/My Drive/101/results/' + clf + '/'\n \n results = []\n matrix = np.zeros(36)\n methods = ['sgd', 'sgdn', 'adam', 'sgd_hd', 'sgdn_hd', 'adam_hd']\n for m in methods:\n for i, s in enumerate(subfolders):\n file = DATA_FOLDER + s + '/' + m + '.csv'\n with open(file, 'r') as f:\n loss = list(csv.reader(f))[-1][4] #training loss\n matrix[i] = np.round(float(loss),3)\n results.append(matrix.reshape(6,-1))\n matrix = np.zeros(36)\n \n all_results.append(results)\n \n", "logreg\nmlp\nbetter_cnn\n" ], [ "import seaborn as sns\nimport matplotlib.pyplot as plt\n\nf, ax = plt.subplots(2,3, figsize=(14,7))\nk=0\nf.suptitle(r'Gridsearch for every ${\\alpha}_{0}$ and $\\beta$ with Logistic Regression on MNIST')\nfor i in range(2):\n for j in range(3):\n ax[i,j].set_title(methods[k])\n f = sns.heatmap(all_results[0][k], annot=True ,cmap=\"YlGnBu\",cbar=True, ax=ax[i,j])\n if k > 2:\n f = sns.heatmap(all_results[0][k], annot=True ,cmap=\"YlGnBu\",cbar=False, ax=ax[i,j], \n mask=np.round(all_results[0][k],2) <= np.round(all_results[0][k-3],2) , annot_kws={\"color\": \"red\"})\n k+=1\n f.set_xticklabels([1e-6, 1e-5, 1e-4, 1e-3, 1e-2, 1e-1],fontsize='small')\n f.set_yticklabels([1e-1, 1e-2, 1e-3, 1e-4, 1e-5, 1e-6],fontsize='small')\n if i==1:\n f.set_xlabel(r'$\\beta_0$')\n f.set_ylabel(r'$\\alpha_0$')\nplt.show()", "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:13: RuntimeWarning: invalid value encountered in less_equal\n del sys.path[0]\n" ], [ "import seaborn as sns\nimport matplotlib.pyplot as plt\n\nf, ax = plt.subplots(2,3, figsize=(14,7))\nk=0\nf.suptitle(r'Gridsearch for every ${\\alpha}_{0}$ and $\\beta$ with MLP on MNIST')\nfor i in range(2):\n for j in range(3):\n ax[i,j].set_title(methods[k])\n f = sns.heatmap(all_results[1][k], annot=True ,cmap=\"YlGnBu\",cbar=True, ax=ax[i,j])\n if k > 2:\n f = sns.heatmap(all_results[1][k], annot=True ,cmap=\"YlGnBu\",cbar=False, ax=ax[i,j], \n mask=np.round(all_results[1][k],2) <= np.round(all_results[1][k-3],2) , annot_kws={\"color\": \"red\"})\n k+=1\n f.set_xticklabels([1e-6, 1e-5, 1e-4, 1e-3, 1e-2, 1e-1],fontsize='small')\n f.set_yticklabels([1e-1, 1e-2, 1e-3, 1e-4, 1e-5, 1e-6],fontsize='small')\n if i==1:\n f.set_xlabel(r'$\\beta_0$')\n f.set_ylabel(r'$\\alpha_0$') \nplt.show()", "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:13: RuntimeWarning: invalid value encountered in less_equal\n del sys.path[0]\n" ], [ "import seaborn \nimport matplotlib.pyplot as plt\n\nf, ax = plt.subplots(2,3, figsize=(14,7))\nk=0\nf.suptitle(r'Gridsearch for every ${\\alpha}_{0}$ and $\\beta$ with \"\"Better CNN\"\" on MNIST')\nfor i in range(2):\n for j in range(3):\n ax[i,j].set_title(methods[k])\n f = sns.heatmap(all_results[2][k], annot=True ,cmap=\"YlGnBu\",cbar=True, ax=ax[i,j])\n if k > 2:\n f = sns.heatmap(all_results[2][k], annot=True ,cmap=\"YlGnBu\",cbar=False, ax=ax[i,j], \n mask=np.round(all_results[2][k],2) <= np.round(all_results[2][k-3],2) , annot_kws={\"color\": \"red\"})\n k+=1\n f.set_xticklabels([1e-6, 1e-5, 1e-4, 1e-3, 1e-2, 1e-1],fontsize='small')\n f.set_yticklabels([1e-1, 1e-2, 1e-3, 1e-4, 1e-5, 1e-6],fontsize='small')\n f.set_xlabel(r'$\\beta_0$')\n f.set_ylabel(r'$\\alpha_0$') \nplt.show()", "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:13: RuntimeWarning: invalid value encountered in less_equal\n del sys.path[0]\n" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "By now basically everyone ([here](http://datacolada.org/2014/06/04/23-ceiling-effects-and-replications/?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+DataColada+%28Data+Colada+Feed%29), [here](http://yorl.tumblr.com/post/87428392426/ceiling-effects), [here](http://www.talyarkoni.org/blog/2014/06/01/there-is-no-ceiling-effect-in-johnson-cheung-donnellan-2014/), [here](http://pigee.wordpress.com/2014/05/24/additional-reflections-on-ceiling-effects-in-recent-replication-research/) and [here](http://www.nicebread.de/reanalyzing-the-schnalljohnson-cleanliness-data-sets-new-insights-from-bayesian-and-robust-approaches/), and there is likely even more out there) who writes a blog and knows how to do a statistical analysis has analysed data from a recent replication study and from the original study (data repository is here). \n\nThe study of two experiments. Let's focus on Experiment 1 here. The experiment consists of a treatment and control group. The performance is measured by six likert-scale items. The scale has 9 levels. All responses are averaged together and we obtain a single composite score for each group. We are interested whether the treatment works, which would show up as a positive difference between the score of the treatment and the control group. Replication study did the same with more subjects.\n\nLet's perform the original analysis to see the results and why this dataset is so \"popular\".", "_____no_output_____" ] ], [ [ "%pylab inline\nimport pystan\nfrom matustools.matusplotlib import *\nfrom scipy import stats", "Populating the interactive namespace from numpy and matplotlib\n" ], [ "il=['dog','trolley','wallet','plane','resume','kitten','mean score','median score']\nD=np.loadtxt('schnallstudy1.csv',delimiter=',')\nD[:,1]=1-D[:,1]\nDtemp=np.zeros((D.shape[0],D.shape[1]+1))\nDtemp[:,:-1]=D\nDtemp[:,-1]=np.median(D[:,2:-2],axis=1)\nD=Dtemp\nDS=D[D[:,0]==0,1:]\nDR=D[D[:,0]==1,1:]\nDS.shape", "_____no_output_____" ], [ "def plotCIttest1(y,x=0,alpha=0.05):\n m=y.mean();df=y.size-1\n se=y.std()/y.size**0.5\n cil=stats.t.ppf(alpha/2.,df)*se\n cii=stats.t.ppf(0.25,df)*se\n out=[m,m-cil,m+cil,m-cii,m+cii]\n _errorbar(out,x=x,clr='k')\n return out\n \ndef plotCIttest2(y1,y2,x=0,alpha=0.05):\n n1=float(y1.size);n2=float(y2.size);\n v1=y1.var();v2=y2.var()\n m=y2.mean()-y1.mean()\n s12=(((n1-1)*v1+(n2-1)*v2)/(n1+n2-2))**0.5\n se=s12*(1/n1+1/n2)**0.5\n df= (v1/n1+v2/n2)**2 / ( (v1/n1)**2/(n1-1)+(v2/n2)**2/(n2-1)) \n cil=stats.t.ppf(alpha/2.,df)*se\n cii=stats.t.ppf(0.25,df)*se\n out=[m,m-cil,m+cil,m-cii,m+cii]\n _errorbar(out,x=x)\n return out\n\nplt.figure(figsize=(4,3))\ndts=[DS[DS[:,0]==0,-2],DS[DS[:,0]==1,-2],\n DR[DR[:,0]==0,-2],DR[DR[:,0]==1,-2]]\nfor k in range(len(dts)):\n plotCIttest1(dts[k],x=k)\nplt.grid(False,axis='x')\nax=plt.gca()\nax.set_xticks(range(len(dts)))\nax.set_xticklabels(['OC','OT','RC','RT'])\nplt.xlim([-0.5,len(dts)-0.5])\nplt.figure(figsize=(4,3))\nplotCIttest2(dts[0],dts[1],x=0,alpha=0.1)\nplotCIttest2(dts[2],dts[3],x=1,alpha=0.1)\nax=plt.gca()\nax.set_xticks([0,1])\nax.set_xticklabels(['OT-OC','RT-RC'])\nplt.grid(False,axis='x')\nplt.xlim([-0.5,1.5]);", "/usr/local/lib/python2.7/dist-packages/matplotlib-1.3.1-py2.7-linux-i686.egg/matplotlib/font_manager.py:1236: UserWarning: findfont: Font family ['Arial'] not found. Falling back to Bitstream Vera Sans\n (prop.get_family(), self.defaultFamily[fontext]))\n/usr/local/lib/python2.7/dist-packages/matplotlib-1.3.1-py2.7-linux-i686.egg/matplotlib/figure.py:1595: UserWarning: This figure includes Axes that are not compatible with tight_layout, so its results might be incorrect.\n warnings.warn(\"This figure includes Axes that are not \"\n" ] ], [ [ "Legend: OC - original study, control group; OT - original study, treatment group; RC - replication study, control group; RT - replication study, treatment group; \n\nIn the original study the difference between the treatment and control is significantly greater than zero. In the replication, it is not. However the ratings in the replication are higher overall. The author of the original study therefore raised a concern that no difference was obtained in replication because of ceiling effects. \n\nHow do we show that there are ceiling efects in the replication? The authors and bloggers presented various arguments that support some conclusion (mostly that there are no ceiling effects). Ultimately ceiling effects are a matter of degree and since no one knows how to quantify them the whole discussion of the replication's validity is heading into an inferential limbo. \n\nMy point here is that if the analysis computed the proper effect size - the causal effect size, we would avoid these kinds of arguments and discussions.", "_____no_output_____" ] ], [ [ "def plotComparison(A,B,stan=False):\n plt.figure(figsize=(8,16))\n cl=['control','treatment']\n x=np.arange(11)-0.5\n if not stan:assert A.shape[1]==B.shape[1]\n for i in range(A.shape[1]-1):\n for cond in range(2):\n plt.subplot(A.shape[1]-1,2,2*i+cond+1)\n a=np.histogram(A[A[:,0]==cond,1+i],bins=x, normed=True)\n plt.barh(x[:-1],-a[0],ec='w',height=1)\n if stan: a=[B[:,i,cond]]\n else: a=np.histogram(B[B[:,0]==cond,1+i],bins=x, normed=True)\n plt.barh(x[:-1],a[0],ec='w',fc='g',height=1)\n #plt.hist(DS[:,2+i],bins=np.arange(11)-0.5,normed=True,rwidth=0.5)\n plt.xlim([-0.7,0.7]);plt.gca().set_yticks(range(10))\n plt.ylim([-1,10]);#plt.grid(b=False,axis='y')\n if not i: plt.title('condition: '+cl[cond])\n if not cond: plt.ylabel(il[i],size=12)\n if not i and not cond: plt.legend(['original','replication'],loc=4);\nplotComparison(DS,DR)", "_____no_output_____" ], [ "model='''\ndata {\n int<lower=2> K;\n int<lower=0> N;\n int<lower=1> M;\n int<lower=1,upper=K> y[N,M];\n int x[N];\n}\nparameters {\n real beta[M];\n ordered[K-1] c[M];\n}\ntransformed parameters{\n real pt[M,K-1]; real pc[M,K-1];\n for (m in 1:M){\n for (k in 1:(K-1)){\n pt[m,k] <- inv_logit(beta[m]-c[m][k]);\n pc[m,k] <- inv_logit(-c[m][k]);\n}}}\nmodel {\nfor (m in 1:M){\n for (k in 1:(K-1)) c[m][k]~ uniform(-100,100);\n for (n in 1:N) y[n,m] ~ ordered_logistic(x[n] * beta[m], c[m]);\n}}\n'''\nsm1=pystan.StanModel(model_code=model)", "INFO:pystan:COMPILING THE C++ CODE FOR MODEL anon_model_0377c9bb8f1b5b3a8a8d14511c81c630 NOW.\n" ], [ "dat = {'y':np.int32(DS[:,1:7])+1,'x':np.int32(DS[:,0]),'N':DS.shape[0] ,'K':10,'M':6}\nfit = sm1.sampling(data=dat,iter=1000, chains=4,thin=2,warmup=500,njobs=2,seed=4)\nprint fit", "Inference for Stan model: anon_model_0377c9bb8f1b5b3a8a8d14511c81c630.\n4 chains, each with iter=1000; warmup=500; thin=2; \npost-warmup draws per chain=250, total post-warmup draws=1000.\n\n mean se_mean sd 2.5% 25% 50% 75% 97.5% n_eff Rhat\nbeta[0] 0.9 0.1 0.6 -0.3 0.5 0.8 1.3 2.1 124.0 1.0\nbeta[1] 0.7 0.0 0.5 -0.3 0.3 0.7 1.1 1.9 151.0 1.0\nbeta[2] 0.4 0.1 0.6 -0.7 -0.0 0.4 0.8 1.5 133.0 1.0\nbeta[3] 0.4 0.1 0.6 -0.9 0.0 0.4 0.8 1.6 99.0 1.0\nbeta[4] 0.8 0.1 0.6 -0.4 0.4 0.8 1.2 2.0 128.0 1.0\nbeta[5] 1.6 0.1 0.7 0.2 1.1 1.6 2.1 2.9 145.0 1.0\nc[0,0] -54.7 3.0 26.8-96.2-79.8-54.2-33.5 -7.4 79.0 1.0\nc[1,0] -1.1 0.0 0.5 -2.1 -1.4 -1.1 -0.8 -0.2 117.0 1.0\nc[2,0] -3.6 0.1 0.9 -6.1 -4.1 -3.6 -3.0 -2.1 80.0 1.0\nc[3,0] -5.2 0.2 1.5 -8.9 -6.1 -5.1 -4.2 -2.9 82.0 1.0\nc[4,0] -69.3 2.4 21.9-99.1-88.7-72.7-52.1 -24.4 84.0 1.0\nc[5,0] -69.9 2.7 21.2-98.9-87.4-73.0-55.3 -23.3 61.0 1.1\nc[0,1] -3.6 0.2 1.0 -6.1 -4.1 -3.4 -2.9 -2.0 42.0 1.1\nc[1,1] -0.2 0.0 0.4 -1.0 -0.4 -0.2 0.1 0.6 152.0 1.0\nc[2,1] -2.8 0.1 0.7 -4.2 -3.3 -2.8 -2.4 -1.6 153.0 1.0\nc[3,1] -3.7 0.1 0.9 -5.8 -4.3 -3.6 -3.0 -2.2 98.0 1.0\nc[4,1] -32.0 2.4 22.4-83.5-48.3-24.9-13.2 -5.0 91.0 1.1\nc[5,1] -35.2 2.5 22.6-83.7-50.3-31.9-15.4 -5.7 82.0 1.0\nc[0,2] -3.1 0.1 0.9 -5.0 -3.6 -3.0 -2.4 -1.6 45.0 1.1\nc[1,2] 0.4 0.0 0.4 -0.4 0.2 0.4 0.7 1.3 140.0 1.0\nc[2,2] -2.0 0.0 0.5 -3.2 -2.3 -2.0 -1.6 -1.1 164.0 1.0\nc[3,2] -2.8 0.1 0.7 -4.3 -3.3 -2.8 -2.3 -1.5 94.0 1.0\nc[4,2] -2.5 0.1 0.7 -4.0 -2.9 -2.4 -2.0 -1.2 90.0 1.0\nc[5,2] -4.3 0.1 1.3 -7.5 -5.1 -4.0 -3.4 -2.4 179.0 1.0\nc[0,3] -1.2 0.0 0.5 -2.2 -1.6 -1.2 -0.9 -0.2 145.0 1.0\nc[1,3] 2.1 0.0 0.5 1.1 1.7 2.1 2.4 3.2 162.0 1.0\nc[2,3] -0.8 0.0 0.4 -1.7 -1.1 -0.8 -0.5 0.0 175.0 1.0\nc[3,3] -1.8 0.0 0.6 -2.9 -2.2 -1.8 -1.4 -0.8 135.0 1.0\nc[4,3] -0.1 0.0 0.5 -1.0 -0.4 -0.1 0.2 0.8 170.0 1.0\nc[5,3] -2.1 0.0 0.6 -3.6 -2.5 -2.0 -1.7 -1.0 185.0 1.0\nc[0,4] -0.4 0.0 0.5 -1.4 -0.7 -0.4 -0.1 0.4 131.0 1.0\nc[1,4] 2.6 0.0 0.6 1.5 2.2 2.6 3.1 4.0 151.0 1.0\nc[2,4] -0.4 0.0 0.4 -1.2 -0.6 -0.3 -0.1 0.4 202.0 1.0\nc[3,4] -1.0 0.0 0.5 -2.0 -1.4 -1.0 -0.7 -0.1 134.0 1.0\nc[4,4] 0.4 0.0 0.5 -0.5 0.1 0.3 0.7 1.2 170.0 1.0\nc[5,4] -1.6 0.0 0.6 -2.9 -2.0 -1.6 -1.2 -0.6 216.0 1.0\nc[0,5] -0.2 0.0 0.4 -1.1 -0.5 -0.2 0.1 0.6 162.0 1.0\nc[1,5] 3.6 0.1 0.8 2.2 3.0 3.5 4.0 5.2 175.0 1.0\nc[2,5] 0.6 0.0 0.4 -0.2 0.3 0.6 0.9 1.5 157.0 1.0\nc[3,5] -0.6 0.0 0.5 -1.5 -0.9 -0.6 -0.2 0.4 151.0 1.0\nc[4,5] 0.6 0.0 0.5 -0.2 0.4 0.6 0.9 1.5 214.0 1.0\nc[5,5] -0.8 0.0 0.5 -1.8 -1.1 -0.8 -0.5 0.1 216.0 1.0\nc[0,6] 0.6 0.0 0.4 -0.2 0.3 0.6 0.9 1.4 222.0 1.0\nc[1,6] 5.4 0.1 1.4 3.2 4.5 5.3 6.2 8.6 150.0 1.0\nc[2,6] 1.1 0.0 0.4 0.3 0.8 1.1 1.4 2.0 189.0 1.0\nc[3,6] 0.2 0.0 0.5 -0.8 -0.1 0.2 0.5 1.1 165.0 1.0\nc[4,6] 1.6 0.0 0.5 0.6 1.3 1.6 2.0 2.7 128.0 1.0\nc[5,6] -0.4 0.0 0.4 -1.3 -0.7 -0.4 -0.1 0.5 211.0 1.0\nc[0,7] 1.0 0.0 0.4 0.2 0.7 1.0 1.3 1.9 170.0 1.0\nc[1,7] 6.5 0.1 1.7 3.8 5.3 6.2 7.5 10.5 152.0 1.0\nc[2,7] 1.3 0.0 0.5 0.5 1.1 1.3 1.6 2.2 195.0 1.0\nc[3,7] 0.8 0.0 0.5 -0.2 0.5 0.8 1.1 1.8 149.0 1.0\nc[4,7] 2.2 0.0 0.6 1.1 1.8 2.2 2.6 3.5 153.0 1.0\nc[5,7] -0.0 0.0 0.4 -0.9 -0.3 -0.0 0.2 0.9 206.0 1.0\nc[0,8] 1.9 0.0 0.5 1.0 1.6 1.9 2.2 3.0 154.0 1.0\nc[1,8] 7.5 0.1 1.9 4.5 6.1 7.1 8.6 11.9 159.0 1.0\nc[2,8] 2.3 0.0 0.5 1.2 1.9 2.2 2.6 3.4 213.0 1.0\nc[3,8] 1.5 0.0 0.5 0.4 1.1 1.5 1.8 2.5 161.0 1.0\nc[4,8] 2.8 0.1 0.7 1.6 2.3 2.7 3.2 4.2 156.0 1.0\nc[5,8] 0.6 0.0 0.5 -0.3 0.3 0.6 0.9 1.6 172.0 1.0\npt[0,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 142.0 1.0\npt[1,0] 0.8 0.0 0.1 0.7 0.8 0.9 0.9 0.9 174.0 1.0\npt[2,0] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 147.0 1.0\npt[3,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 82.0 1.0\npt[4,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 306.0 1.0\npt[5,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 325.0 1.0\npt[0,1] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 110.0 1.0\npt[1,1] 0.7 0.0 0.1 0.5 0.6 0.7 0.8 0.9 185.0 1.0\npt[2,1] 1.0 0.0 0.0 0.9 0.9 1.0 1.0 1.0 150.0 1.0\npt[3,1] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 150.0 1.0\npt[4,1] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 181.0 1.0\npt[5,1] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 144.0 1.0\npt[0,2] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 146.0 1.0\npt[1,2] 0.6 0.0 0.1 0.4 0.5 0.6 0.6 0.8 234.0 1.0\npt[2,2] 0.9 0.0 0.1 0.8 0.9 0.9 0.9 1.0 155.0 1.0\npt[3,2] 1.0 0.0 0.0 0.9 0.9 1.0 1.0 1.0 169.0 1.0\npt[4,2] 1.0 0.0 0.0 0.9 0.9 1.0 1.0 1.0 76.0 1.1\npt[5,2] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 216.0 1.0\npt[0,3] 0.9 0.0 0.1 0.7 0.9 0.9 0.9 1.0 169.0 1.0\npt[1,3] 0.2 0.0 0.1 0.1 0.2 0.2 0.3 0.4 165.0 1.0\npt[2,3] 0.8 0.0 0.1 0.5 0.7 0.8 0.8 0.9 174.0 1.0\npt[3,3] 0.9 0.0 0.1 0.8 0.9 0.9 0.9 1.0 160.0 1.0\npt[4,3] 0.7 0.0 0.1 0.5 0.6 0.7 0.8 0.9 157.0 1.0\npt[5,3] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 113.0 1.0\npt[0,4] 0.8 0.0 0.1 0.6 0.7 0.8 0.8 0.9 196.0 1.0\npt[1,4] 0.1 0.0 0.1 0.0 0.1 0.1 0.2 0.3 158.0 1.0\npt[2,4] 0.7 0.0 0.1 0.4 0.6 0.7 0.8 0.9 171.0 1.0\npt[3,4] 0.8 0.0 0.1 0.6 0.7 0.8 0.9 0.9 117.0 1.0\npt[4,4] 0.6 0.0 0.1 0.4 0.5 0.6 0.7 0.8 162.0 1.0\npt[5,4] 1.0 0.0 0.0 0.9 0.9 1.0 1.0 1.0 110.0 1.0\npt[0,5] 0.7 0.0 0.1 0.6 0.7 0.7 0.8 0.9 199.0 1.0\npt[1,5] 0.1 0.0 0.0 0.0 0.0 0.1 0.1 0.2 178.0 1.0\npt[2,5] 0.5 0.0 0.1 0.3 0.4 0.5 0.5 0.7 189.0 1.0\npt[3,5] 0.7 0.0 0.1 0.5 0.7 0.7 0.8 0.9 129.0 1.0\npt[4,5] 0.5 0.0 0.1 0.3 0.5 0.5 0.6 0.7 178.0 1.0\npt[5,5] 0.9 0.0 0.0 0.8 0.9 0.9 0.9 1.0 163.0 1.0\npt[0,6] 0.6 0.0 0.1 0.3 0.5 0.6 0.7 0.8 157.0 1.0\npt[1,6] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 170.0 1.0\npt[2,6] 0.3 0.0 0.1 0.2 0.3 0.3 0.4 0.5 194.0 1.0\npt[3,6] 0.6 0.0 0.1 0.3 0.5 0.6 0.6 0.7 116.0 1.0\npt[4,6] 0.3 0.0 0.1 0.1 0.2 0.3 0.4 0.5 174.0 1.0\npt[5,6] 0.9 0.0 0.1 0.7 0.8 0.9 0.9 1.0 135.0 1.0\npt[0,7] 0.5 0.0 0.1 0.3 0.4 0.5 0.5 0.7 178.0 1.0\npt[1,7] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 192.0 1.0\npt[2,7] 0.3 0.0 0.1 0.1 0.2 0.3 0.4 0.5 186.0 1.0\npt[3,7] 0.4 0.0 0.1 0.2 0.3 0.4 0.5 0.6 149.0 1.0\npt[4,7] 0.2 0.0 0.1 0.1 0.1 0.2 0.3 0.4 205.0 1.0\npt[5,7] 0.8 0.0 0.1 0.6 0.8 0.8 0.9 0.9 164.0 1.0\npt[0,8] 0.3 0.0 0.1 0.1 0.2 0.3 0.3 0.5 166.0 1.0\npt[1,8] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 208.0 1.0\npt[2,8] 0.2 0.0 0.1 0.0 0.1 0.1 0.2 0.3 146.0 1.0\npt[3,8] 0.3 0.0 0.1 0.1 0.2 0.3 0.3 0.5 120.0 1.0\npt[4,8] 0.1 0.0 0.1 0.0 0.1 0.1 0.2 0.3 232.0 1.0\npt[5,8] 0.7 0.0 0.1 0.5 0.7 0.7 0.8 0.9 196.0 1.0\npc[0,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 152.0 1.0\npc[1,0] 0.7 0.0 0.1 0.6 0.7 0.7 0.8 0.9 123.0 1.0\npc[2,0] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 137.0 1.0\npc[3,0] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 59.0 1.1\npc[4,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 303.0 1.0\npc[5,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 327.0 1.0\npc[0,1] 1.0 0.0 0.0 0.9 0.9 1.0 1.0 1.0 86.0 1.0\npc[1,1] 0.5 0.0 0.1 0.4 0.5 0.5 0.6 0.7 152.0 1.0\npc[2,1] 0.9 0.0 0.0 0.8 0.9 0.9 1.0 1.0 153.0 1.0\npc[3,1] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 97.0 1.0\npc[4,1] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 163.0 1.0\npc[5,1] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 172.0 1.0\npc[0,2] 0.9 0.0 0.0 0.8 0.9 1.0 1.0 1.0 86.0 1.0\npc[1,2] 0.4 0.0 0.1 0.2 0.3 0.4 0.5 0.6 139.0 1.0\npc[2,2] 0.9 0.0 0.1 0.7 0.8 0.9 0.9 1.0 163.0 1.0\npc[3,2] 0.9 0.0 0.0 0.8 0.9 0.9 1.0 1.0 87.0 1.0\npc[4,2] 0.9 0.0 0.1 0.8 0.9 0.9 0.9 1.0 95.0 1.0\npc[5,2] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 208.0 1.0\npc[0,3] 0.8 0.0 0.1 0.6 0.7 0.8 0.8 0.9 136.0 1.0\npc[1,3] 0.1 0.0 0.1 0.0 0.1 0.1 0.2 0.3 160.0 1.0\npc[2,3] 0.7 0.0 0.1 0.5 0.6 0.7 0.7 0.8 177.0 1.0\npc[3,3] 0.8 0.0 0.1 0.7 0.8 0.9 0.9 0.9 145.0 1.0\npc[4,3] 0.5 0.0 0.1 0.3 0.5 0.5 0.6 0.7 169.0 1.0\npc[5,3] 0.9 0.0 0.1 0.7 0.8 0.9 0.9 1.0 187.0 1.0\npc[0,4] 0.6 0.0 0.1 0.4 0.5 0.6 0.7 0.8 130.0 1.0\npc[1,4] 0.1 0.0 0.0 0.0 0.0 0.1 0.1 0.2 159.0 1.0\npc[2,4] 0.6 0.0 0.1 0.4 0.5 0.6 0.7 0.8 202.0 1.0\npc[3,4] 0.7 0.0 0.1 0.5 0.7 0.7 0.8 0.9 135.0 1.0\npc[4,4] 0.4 0.0 0.1 0.2 0.3 0.4 0.5 0.6 212.0 1.0\npc[5,4] 0.8 0.0 0.1 0.7 0.8 0.8 0.9 0.9 221.0 1.0\npc[0,5] 0.6 0.0 0.1 0.3 0.5 0.6 0.6 0.7 160.0 1.0\npc[1,5] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 171.0 1.0\npc[2,5] 0.4 0.0 0.1 0.2 0.3 0.4 0.4 0.6 168.0 1.0\npc[3,5] 0.6 0.0 0.1 0.4 0.6 0.6 0.7 0.8 156.0 1.0\npc[4,5] 0.3 0.0 0.1 0.2 0.3 0.3 0.4 0.6 214.0 1.0\npc[5,5] 0.7 0.0 0.1 0.5 0.6 0.7 0.7 0.9 228.0 1.0\npc[0,6] 0.4 0.0 0.1 0.2 0.3 0.4 0.4 0.6 224.0 1.0\npc[1,6] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 190.0 1.0\npc[2,6] 0.3 0.0 0.1 0.1 0.2 0.2 0.3 0.4 191.0 1.0\npc[3,6] 0.5 0.0 0.1 0.3 0.4 0.5 0.5 0.7 166.0 1.0\npc[4,6] 0.2 0.0 0.1 0.1 0.1 0.2 0.2 0.3 133.0 1.0\npc[5,6] 0.6 0.0 0.1 0.4 0.5 0.6 0.7 0.8 212.0 1.0\npc[0,7] 0.3 0.0 0.1 0.1 0.2 0.3 0.3 0.5 184.0 1.0\npc[1,7] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 189.0 1.0\npc[2,7] 0.2 0.0 0.1 0.1 0.2 0.2 0.3 0.4 199.0 1.0\npc[3,7] 0.3 0.0 0.1 0.1 0.2 0.3 0.4 0.5 151.0 1.0\npc[4,7] 0.1 0.0 0.1 0.0 0.1 0.1 0.1 0.2 161.0 1.0\npc[5,7] 0.5 0.0 0.1 0.3 0.4 0.5 0.6 0.7 205.0 1.0\npc[0,8] 0.1 0.0 0.1 0.0 0.1 0.1 0.2 0.3 147.0 1.0\npc[1,8] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 196.0 1.0\npc[2,8] 0.1 0.0 0.1 0.0 0.1 0.1 0.1 0.2 206.0 1.0\npc[3,8] 0.2 0.0 0.1 0.1 0.1 0.2 0.2 0.4 164.0 1.0\npc[4,8] 0.1 0.0 0.0 0.0 0.0 0.1 0.1 0.2 166.0 1.0\npc[5,8] 0.4 0.0 0.1 0.2 0.3 0.3 0.4 0.6 176.0 1.0\nlp__ -478.9 0.5 5.4-489.8-482.4-478.7-475.3-468.9 136.0 1.0\n\nSamples were drawn using NUTS(diag_e) at Sat Jun 14 01:33:17 2014.\nFor each parameter, n_eff is a crude measure of effective sample size,\nand Rhat is the potential scale reduction factor on split chains (at \nconvergence, Rhat=1).\n" ], [ "pt=fit.extract()['pt']\npc=fit.extract()['pc']\nDP=np.zeros((pt.shape[2]+2,pt.shape[1],2))\nDP[0,:,:]=1\nDP[1:-1,:,:]=np.array([pc,pt]).T.mean(2)\nDP=-np.diff(DP,axis=0)\nplotComparison(DS[:,:7],DP,stan=True)", "_____no_output_____" ], [ "model='''\ndata {\n int<lower=2> K;\n int<lower=0> N;\n int<lower=1> M;\n int<lower=1,upper=K> y[N,M];\n int x[N];\n}\nparameters {\n real beta;\n ordered[K-1] c[M];\n}\ntransformed parameters{\n real pt[M,K-1]; real pc[M,K-1];\n for (m in 1:M){\n for (k in 1:(K-1)){\n pt[m,k] <- inv_logit(beta-c[m][k]);\n pc[m,k] <- inv_logit(-c[m][k]);\n}}}\nmodel {\nfor (m in 1:M){\n for (k in 1:(K-1)) c[m][k]~ uniform(-100,100);\n for (n in 1:N) y[n,m] ~ ordered_logistic(x[n] * beta, c[m]);\n}}\n'''\nsm2=pystan.StanModel(model_code=model)", "_____no_output_____" ], [ "dat = {'y':np.int32(DS[:,1:7])+1,'x':np.int32(DS[:,0]),'N':DS.shape[0] ,'K':10,'M':6}\nfit2 = sm2.sampling(data=dat,iter=1000, chains=4,thin=2,warmup=500,njobs=2,seed=4)\nprint fit2\nsaveStanFit(fit2,'fit2')", "Inference for Stan model: anon_model_7216e04af35b427b9a410fdc88ae5601.\n4 chains, each with iter=1000; warmup=500; thin=2; \npost-warmup draws per chain=250, total post-warmup draws=1000.\n\n mean se_mean sd 2.5% 25% 50% 75% 97.5% n_eff Rhat\nbeta 0.7 0.0 0.2 0.3 0.6 0.7 0.9 1.2 132.0 1.0\nc[0,0] -53.7 4.2 26.8-97.1-75.8-55.2-31.5 -8.7 40.0 1.1\nc[1,0] -1.1 0.0 0.4 -1.8 -1.3 -1.1 -0.8 -0.4 103.0 1.0\nc[2,0] -3.6 0.1 1.0 -5.7 -4.2 -3.5 -2.9 -2.0 79.0 1.0\nc[3,0] -5.2 0.2 1.4 -8.5 -5.9 -4.9 -4.1 -3.0 75.0 1.0\nc[4,0] -68.0 2.2 22.1-99.0-86.7-71.2-52.1 -21.5 106.0 1.0\nc[5,0] -71.0 2.3 21.8-99.1-89.1-75.7-54.6 -23.0 88.0 1.0\nc[0,1] -3.5 0.1 0.9 -5.4 -4.1 -3.4 -2.9 -2.0 172.0 1.0\nc[1,1] -0.1 0.0 0.3 -0.8 -0.4 -0.1 0.1 0.5 152.0 1.0\nc[2,1] -2.7 0.1 0.7 -4.2 -3.1 -2.7 -2.2 -1.5 117.0 1.0\nc[3,1] -3.7 0.1 1.1 -6.8 -4.1 -3.5 -3.0 -2.1 64.0 1.1\nc[4,1] -36.2 2.0 21.4-81.5-51.8-33.3-18.3 -5.3 114.0 1.0\nc[5,1] -39.9 2.5 22.3-86.7-56.5-36.6-21.2 -7.3 79.0 1.0\nc[0,2] -3.0 0.1 0.8 -4.8 -3.4 -2.9 -2.5 -1.7 187.0 1.0\nc[1,2] 0.4 0.0 0.3 -0.3 0.2 0.4 0.6 1.1 137.0 1.0\nc[2,2] -1.9 0.0 0.5 -2.9 -2.2 -1.8 -1.5 -0.9 145.0 1.0\nc[3,2] -2.7 0.1 0.7 -4.2 -3.1 -2.6 -2.2 -1.4 109.0 1.0\nc[4,2] -2.5 0.1 0.7 -4.2 -2.9 -2.4 -2.0 -1.2 137.0 1.0\nc[5,2] -4.7 0.1 1.4 -8.1 -5.5 -4.5 -3.7 -2.7 89.0 1.0\nc[0,3] -1.3 0.0 0.4 -2.1 -1.5 -1.3 -1.0 -0.5 153.0 1.0\nc[1,3] 2.0 0.0 0.5 1.2 1.7 2.0 2.3 2.9 159.0 1.0\nc[2,3] -0.7 0.0 0.4 -1.5 -0.9 -0.7 -0.4 0.1 220.0 1.0\nc[3,3] -1.7 0.0 0.5 -2.7 -2.0 -1.7 -1.4 -0.8 154.0 1.0\nc[4,3] -0.1 0.0 0.4 -0.8 -0.4 -0.1 0.1 0.6 221.0 1.0\nc[5,3] -2.4 0.0 0.6 -3.7 -2.8 -2.4 -2.0 -1.4 159.0 1.0\nc[0,4] -0.6 0.0 0.4 -1.3 -0.8 -0.6 -0.3 0.2 152.0 1.0\nc[1,4] 2.6 0.0 0.6 1.6 2.2 2.5 2.9 3.9 148.0 1.0\nc[2,4] -0.2 0.0 0.4 -1.0 -0.5 -0.2 -0.0 0.5 221.0 1.0\nc[3,4] -0.9 0.0 0.4 -1.7 -1.2 -0.9 -0.6 -0.1 78.0 1.0\nc[4,4] 0.4 0.0 0.4 -0.3 0.1 0.3 0.6 1.1 234.0 1.0\nc[5,4] -1.9 0.0 0.5 -3.1 -2.3 -1.9 -1.6 -1.0 196.0 1.0\nc[0,5] -0.3 0.0 0.4 -1.1 -0.6 -0.3 -0.1 0.4 200.0 1.0\nc[1,5] 3.4 0.1 0.7 2.1 2.9 3.4 3.9 4.9 139.0 1.0\nc[2,5] 0.7 0.0 0.4 0.0 0.5 0.7 1.0 1.4 219.0 1.0\nc[3,5] -0.4 0.0 0.4 -1.2 -0.7 -0.4 -0.2 0.3 134.0 1.0\nc[4,5] 0.6 0.0 0.4 -0.0 0.4 0.6 0.9 1.4 267.0 1.0\nc[5,5] -1.1 0.0 0.4 -2.0 -1.4 -1.1 -0.8 -0.3 257.0 1.0\nc[0,6] 0.5 0.0 0.3 -0.2 0.2 0.5 0.7 1.2 246.0 1.0\nc[1,6] 5.3 0.1 1.4 3.1 4.3 5.1 6.1 8.6 153.0 1.0\nc[2,6] 1.3 0.0 0.4 0.5 1.0 1.2 1.5 2.0 223.0 1.0\nc[3,6] 0.3 0.0 0.3 -0.3 0.1 0.3 0.5 1.0 264.0 1.0\nc[4,6] 1.6 0.0 0.4 0.8 1.3 1.6 1.9 2.4 253.0 1.0\nc[5,6] -0.7 0.0 0.4 -1.6 -1.0 -0.7 -0.5 -0.0 287.0 1.0\nc[0,7] 0.9 0.0 0.4 0.2 0.6 0.9 1.1 1.5 228.0 1.0\nc[1,7] 6.3 0.2 1.8 3.7 5.0 5.9 7.4 10.7 139.0 1.0\nc[2,7] 1.5 0.0 0.4 0.7 1.2 1.5 1.8 2.3 207.0 1.0\nc[3,7] 0.9 0.0 0.4 0.2 0.7 0.9 1.2 1.7 170.0 1.0\nc[4,7] 2.2 0.0 0.5 1.3 1.9 2.2 2.5 3.2 232.0 1.0\nc[5,7] -0.4 0.0 0.4 -1.1 -0.6 -0.4 -0.1 0.3 306.0 1.0\nc[0,8] 1.8 0.0 0.4 1.0 1.5 1.8 2.1 2.6 218.0 1.0\nc[1,8] 7.4 0.2 2.2 4.0 5.8 7.1 8.8 12.3 125.0 1.0\nc[2,8] 2.4 0.0 0.5 1.4 2.0 2.4 2.7 3.5 166.0 1.0\nc[3,8] 1.6 0.0 0.4 0.8 1.4 1.6 1.9 2.4 151.0 1.0\nc[4,8] 2.8 0.0 0.6 1.7 2.4 2.7 3.1 4.1 204.0 1.0\nc[5,8] 0.2 0.0 0.4 -0.5 -0.0 0.2 0.5 0.9 279.0 1.0\npt[0,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 290.0 1.0\npt[1,0] 0.9 0.0 0.0 0.7 0.8 0.9 0.9 0.9 137.0 1.0\npt[2,0] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 70.0 1.0\npt[3,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 85.0 1.0\npt[4,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 248.0 1.0\npt[5,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 289.0 1.0\npt[0,1] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 137.0 1.0\npt[1,1] 0.7 0.0 0.1 0.6 0.7 0.7 0.8 0.8 197.0 1.0\npt[2,1] 1.0 0.0 0.0 0.9 0.9 1.0 1.0 1.0 92.0 1.0\npt[3,1] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 114.0 1.0\npt[4,1] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 209.0 1.0\npt[5,1] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 202.0 1.0\npt[0,2] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 131.0 1.0\npt[1,2] 0.6 0.0 0.1 0.4 0.5 0.6 0.6 0.7 193.0 1.0\npt[2,2] 0.9 0.0 0.0 0.8 0.9 0.9 0.9 1.0 146.0 1.0\npt[3,2] 1.0 0.0 0.0 0.9 0.9 1.0 1.0 1.0 135.0 1.0\npt[4,2] 1.0 0.0 0.0 0.9 0.9 1.0 1.0 1.0 155.0 1.0\npt[5,2] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 150.0 1.0\npt[0,3] 0.9 0.0 0.0 0.8 0.8 0.9 0.9 0.9 159.0 1.0\npt[1,3] 0.2 0.0 0.1 0.1 0.2 0.2 0.3 0.4 208.0 1.0\npt[2,3] 0.8 0.0 0.1 0.7 0.8 0.8 0.8 0.9 187.0 1.0\npt[3,3] 0.9 0.0 0.0 0.8 0.9 0.9 0.9 1.0 140.0 1.0\npt[4,3] 0.7 0.0 0.1 0.5 0.6 0.7 0.7 0.8 233.0 1.0\npt[5,3] 1.0 0.0 0.0 0.9 0.9 1.0 1.0 1.0 156.0 1.0\npt[0,4] 0.8 0.0 0.1 0.6 0.7 0.8 0.8 0.9 196.0 1.0\npt[1,4] 0.1 0.0 0.1 0.0 0.1 0.1 0.2 0.3 189.0 1.0\npt[2,4] 0.7 0.0 0.1 0.6 0.7 0.7 0.8 0.9 161.0 1.0\npt[3,4] 0.8 0.0 0.1 0.7 0.8 0.8 0.9 0.9 87.0 1.0\npt[4,4] 0.6 0.0 0.1 0.4 0.5 0.6 0.6 0.7 228.0 1.0\npt[5,4] 0.9 0.0 0.0 0.9 0.9 0.9 1.0 1.0 203.0 1.0\npt[0,5] 0.7 0.0 0.1 0.6 0.7 0.7 0.8 0.9 238.0 1.0\npt[1,5] 0.1 0.0 0.0 0.0 0.0 0.1 0.1 0.2 153.0 1.0\npt[2,5] 0.5 0.0 0.1 0.3 0.4 0.5 0.6 0.7 255.0 1.0\npt[3,5] 0.8 0.0 0.1 0.6 0.7 0.8 0.8 0.9 152.0 1.0\npt[4,5] 0.5 0.0 0.1 0.4 0.5 0.5 0.6 0.7 206.0 1.0\npt[5,5] 0.9 0.0 0.1 0.7 0.8 0.9 0.9 0.9 213.0 1.0\npt[0,6] 0.6 0.0 0.1 0.4 0.5 0.6 0.6 0.7 209.0 1.0\npt[1,6] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 187.0 1.0\npt[2,6] 0.4 0.0 0.1 0.2 0.3 0.4 0.4 0.6 251.0 1.0\npt[3,6] 0.6 0.0 0.1 0.4 0.5 0.6 0.7 0.7 256.0 1.0\npt[4,6] 0.3 0.0 0.1 0.2 0.2 0.3 0.4 0.5 233.0 1.0\npt[5,6] 0.8 0.0 0.1 0.7 0.8 0.8 0.9 0.9 239.0 1.0\npt[0,7] 0.5 0.0 0.1 0.3 0.4 0.5 0.5 0.6 193.0 1.0\npt[1,7] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 174.0 1.0\npt[2,7] 0.3 0.0 0.1 0.2 0.3 0.3 0.4 0.5 212.0 1.0\npt[3,7] 0.5 0.0 0.1 0.3 0.4 0.4 0.5 0.6 246.0 1.0\npt[4,7] 0.2 0.0 0.1 0.1 0.1 0.2 0.2 0.4 215.0 1.0\npt[5,7] 0.7 0.0 0.1 0.6 0.7 0.8 0.8 0.9 225.0 1.0\npt[0,8] 0.3 0.0 0.1 0.1 0.2 0.3 0.3 0.4 257.0 1.0\npt[1,8] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 180.0 1.0\npt[2,8] 0.2 0.0 0.1 0.1 0.1 0.2 0.2 0.3 159.0 1.0\npt[3,8] 0.3 0.0 0.1 0.2 0.2 0.3 0.3 0.5 167.0 1.0\npt[4,8] 0.1 0.0 0.1 0.0 0.1 0.1 0.2 0.3 203.0 1.0\npt[5,8] 0.6 0.0 0.1 0.5 0.6 0.6 0.7 0.8 253.0 1.0\npc[0,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 288.0 1.0\npc[1,0] 0.7 0.0 0.1 0.6 0.7 0.7 0.8 0.9 100.0 1.0\npc[2,0] 1.0 0.0 0.0 0.9 0.9 1.0 1.0 1.0 78.0 1.0\npc[3,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 72.0 1.0\npc[4,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 248.0 1.0\npc[5,0] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 280.0 1.0\npc[0,1] 1.0 0.0 0.0 0.9 0.9 1.0 1.0 1.0 159.0 1.0\npc[1,1] 0.5 0.0 0.1 0.4 0.5 0.5 0.6 0.7 152.0 1.0\npc[2,1] 0.9 0.0 0.0 0.8 0.9 0.9 1.0 1.0 106.0 1.0\npc[3,1] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 101.0 1.0\npc[4,1] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 218.0 1.0\npc[5,1] 1.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 194.0 1.0\npc[0,2] 0.9 0.0 0.0 0.8 0.9 0.9 1.0 1.0 161.0 1.0\npc[1,2] 0.4 0.0 0.1 0.3 0.3 0.4 0.5 0.6 135.0 1.0\npc[2,2] 0.9 0.0 0.1 0.7 0.8 0.9 0.9 0.9 149.0 1.0\npc[3,2] 0.9 0.0 0.0 0.8 0.9 0.9 1.0 1.0 123.0 1.0\npc[4,2] 0.9 0.0 0.1 0.8 0.9 0.9 0.9 1.0 161.0 1.0\npc[5,2] 1.0 0.0 0.0 0.9 1.0 1.0 1.0 1.0 122.0 1.0\npc[0,3] 0.8 0.0 0.1 0.6 0.7 0.8 0.8 0.9 154.0 1.0\npc[1,3] 0.1 0.0 0.0 0.1 0.1 0.1 0.2 0.2 167.0 1.0\npc[2,3] 0.7 0.0 0.1 0.5 0.6 0.7 0.7 0.8 223.0 1.0\npc[3,3] 0.8 0.0 0.1 0.7 0.8 0.8 0.9 0.9 148.0 1.0\npc[4,3] 0.5 0.0 0.1 0.4 0.5 0.5 0.6 0.7 221.0 1.0\npc[5,3] 0.9 0.0 0.0 0.8 0.9 0.9 0.9 1.0 159.0 1.0\npc[0,4] 0.6 0.0 0.1 0.5 0.6 0.6 0.7 0.8 159.0 1.0\npc[1,4] 0.1 0.0 0.0 0.0 0.1 0.1 0.1 0.2 180.0 1.0\npc[2,4] 0.6 0.0 0.1 0.4 0.5 0.6 0.6 0.7 220.0 1.0\npc[3,4] 0.7 0.0 0.1 0.5 0.6 0.7 0.8 0.8 78.0 1.0\npc[4,4] 0.4 0.0 0.1 0.3 0.4 0.4 0.5 0.6 231.0 1.0\npc[5,4] 0.9 0.0 0.1 0.7 0.8 0.9 0.9 1.0 193.0 1.0\npc[0,5] 0.6 0.0 0.1 0.4 0.5 0.6 0.6 0.8 201.0 1.0\npc[1,5] 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.1 150.0 1.0\npc[2,5] 0.3 0.0 0.1 0.2 0.3 0.3 0.4 0.5 216.0 1.0\npc[3,5] 0.6 0.0 0.1 0.4 0.5 0.6 0.7 0.8 130.0 1.0\npc[4,5] 0.3 0.0 0.1 0.2 0.3 0.3 0.4 0.5 270.0 1.0\npc[5,5] 0.7 0.0 0.1 0.6 0.7 0.7 0.8 0.9 252.0 1.0\npc[0,6] 0.4 0.0 0.1 0.2 0.3 0.4 0.4 0.6 244.0 1.0\npc[1,6] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 195.0 1.0\npc[2,6] 0.2 0.0 0.1 0.1 0.2 0.2 0.3 0.4 222.0 1.0\npc[3,6] 0.4 0.0 0.1 0.3 0.4 0.4 0.5 0.6 264.0 1.0\npc[4,6] 0.2 0.0 0.1 0.1 0.1 0.2 0.2 0.3 257.0 1.0\npc[5,6] 0.7 0.0 0.1 0.5 0.6 0.7 0.7 0.8 285.0 1.0\npc[0,7] 0.3 0.0 0.1 0.2 0.2 0.3 0.3 0.5 228.0 1.0\npc[1,7] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 182.0 1.0\npc[2,7] 0.2 0.0 0.1 0.1 0.1 0.2 0.2 0.3 208.0 1.0\npc[3,7] 0.3 0.0 0.1 0.2 0.2 0.3 0.3 0.4 166.0 1.0\npc[4,7] 0.1 0.0 0.0 0.0 0.1 0.1 0.1 0.2 236.0 1.0\npc[5,7] 0.6 0.0 0.1 0.4 0.5 0.6 0.6 0.8 306.0 1.0\npc[0,8] 0.2 0.0 0.1 0.1 0.1 0.1 0.2 0.3 222.0 1.0\npc[1,8] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 182.0 1.0\npc[2,8] 0.1 0.0 0.0 0.0 0.1 0.1 0.1 0.2 158.0 1.0\npc[3,8] 0.2 0.0 0.1 0.1 0.1 0.2 0.2 0.3 147.0 1.0\npc[4,8] 0.1 0.0 0.0 0.0 0.0 0.1 0.1 0.1 209.0 1.0\npc[5,8] 0.4 0.0 0.1 0.3 0.4 0.4 0.5 0.6 279.0 1.0\nlp__ -478.0 0.6 6.1-492.6-481.5-477.6-473.6-467.5 109.0 1.0\n\nSamples were drawn using NUTS(diag_e) at Sat Jun 14 10:05:05 2014.\nFor each parameter, n_eff is a crude measure of effective sample size,\nand Rhat is the potential scale reduction factor on split chains (at \nconvergence, Rhat=1).\n" ], [ "w=loadStanFit('fit2')\npt=w['pt']\npc=w['pc']\nDP=np.zeros((pt.shape[2]+2,pt.shape[1],2))\nDP[0,:,:]=1\nDP[1:-1,:,:]=np.array([pc,pt]).T.mean(2)\nDP=-np.diff(DP,axis=0)\nplotComparison(DS[:,:7],DP,stan=True)", "_____no_output_____" ], [ "model='''\ndata {\n int<lower=2> K;\n int<lower=0> N;\n int<lower=1> M;\n int<lower=1,upper=K> y[N,M];\n int x[N];\n}\nparameters {\n // real mb; real<lower=0,upper=100> sb[2];\n vector[2*M-1] bbeta;\n ordered[K-1] c;\n}\ntransformed parameters{\n real pt[M,K-1]; real pc[M,K-1];\n vector[M] beta[2];\n for (m in 1:M){\n if (m==1){beta[1][m]<-0.0; beta[2][m]<-bbeta[2*M-1];}\n else{beta[1][m]<-bbeta[2*(m-1)-1]; beta[2][m]<-bbeta[2*(m-1)];}\n for (k in 1:(K-1)){\n pt[m,k] <- inv_logit(beta[2][m]-c[k]);\n pc[m,k] <- inv_logit(beta[1][m]-c[k]);\n}}}\nmodel {\nfor (k in 1:(K-1)) c[k]~ uniform(-100,100);\n//beta[1]~normal(0.0,sb[1]);\n//beta[2]~normal(mb,sb[2]);\nfor (m in 1:M){\n for (n in 1:N) y[n,m] ~ ordered_logistic(beta[x[n]+1][m], c);\n}}\n'''\nsm3=pystan.StanModel(model_code=model)", "INFO:pystan:COMPILING THE C++ CODE FOR MODEL anon_model_b233c89bb9dd8a0b522a51670536ebd9 NOW.\n" ], [ "dat = {'y':np.int32(DS[:,1:7])+1,'x':np.int32(DS[:,0]),'N':DS.shape[0] ,'K':10,'M':6}\nfit3 = sm3.sampling(data=dat,iter=1000, chains=4,thin=2,warmup=500,njobs=2,seed=4)\n#print fit3\nsaveStanFit(fit3,'fit3')", "_____no_output_____" ], [ "w=loadStanFit('fit3')\npt=w['pt']\npc=w['pc']\nDP=np.zeros((pt.shape[2]+2,pt.shape[1],2))\nDP[0,:,:]=1\nDP[1:-1,:,:]=np.array([pc,pt]).T.mean(2)\nDP=-np.diff(DP,axis=0)\nplotComparison(DS[:,:7],DP,stan=True)", "_____no_output_____" ], [ "model='''\ndata {\n int<lower=2> K;\n int<lower=0> N;\n int<lower=1> M;\n int<lower=1,upper=K> y[N,M];\n int x[N];\n}\nparameters {\n // real mb; real<lower=0,upper=100> sb[2];\n vector[M-1] bbeta;\n real delt;\n ordered[K-1] c;\n}\ntransformed parameters{\n real pt[M,K-1]; real pc[M,K-1];\n vector[M] beta;\n for (m in 1:M){\n if (m==1) beta[m]<-0.0; \n else beta[m]<-bbeta[m-1];\n for (k in 1:(K-1)){\n pt[m,k] <- inv_logit(beta[m]+delt-c[k]);\n pc[m,k] <- inv_logit(beta[m]-c[k]);\n}}}\nmodel {\nfor (k in 1:(K-1)) c[k]~ uniform(-100,100);\nfor (m in 1:M){\n for (n in 1:N) y[n,m] ~ ordered_logistic(beta[m]+delt*x[n], c);\n}}\n'''\nsm4=pystan.StanModel(model_code=model)", "_____no_output_____" ], [ "dat = {'y':np.int32(DS[:,1:7])+1,'x':np.int32(DS[:,0]),'N':DS.shape[0] ,'K':10,'M':6}\nfit4 = sm4.sampling(data=dat,iter=1000, chains=4,thin=2,warmup=500,njobs=2,seed=4)\nprint pystan.misc._print_stanfit(fit4,pars=['delt','bbeta','c'],digits_summary=2)\nsaveStanFit(fit4,'fit4')", "Inference for Stan model: anon_model_ea755999b4327833f97905399d8651f3.\n4 chains, each with iter=1000; warmup=500; thin=2; \npost-warmup draws per chain=250, total post-warmup draws=1000.\n\n mean se_mean sd 2.5% 25% 50% 75% 97.5% n_eff Rhat\ndelt 0.67 0.01 0.24 0.21 0.51 0.67 0.83 1.14 334.0 1.0\nbbeta[0]-3.16 0.03 0.46-4.04-3.47-3.16-2.85 -2.26 240.0 1.0\nbbeta[1]-0.68 0.02 0.4-1.47-0.95-0.68-0.42 0.11 264.0 1.0\nbbeta[2] 0.1 0.02 0.4-0.67-0.19 0.11 0.36 0.94 288.0 1.0\nbbeta[3]-0.91 0.02 0.4-1.67-1.19-0.91-0.63 -0.14 260.0 1.0\nbbeta[4] 1.24 0.02 0.41 0.4 0.97 1.24 1.51 2.02 302.0 1.0\nc[0] -4.22 0.03 0.46-5.14-4.53-4.22-3.92 -3.28 272.0 1.0\nc[1] -3.32 0.03 0.41-4.13 -3.6-3.34-3.02 -2.46 249.0 1.0\nc[2] -2.65 0.02 0.39-3.37-2.91-2.66-2.39 -1.85 269.0 1.0\nc[3] -1.16 0.02 0.33-1.78-1.38-1.17-0.93 -0.51 256.0 1.0\nc[4] -0.68 0.02 0.32-1.28-0.89-0.68-0.47 -0.04 267.0 1.0\nc[5] -0.2 0.02 0.31-0.79-0.42-0.19-0.01 0.47 271.0 1.0\nc[6] 0.48 0.02 0.31-0.08 0.26 0.48 0.68 1.11 280.0 1.0\nc[7] 0.83 0.02 0.31 0.26 0.61 0.84 1.04 1.5 281.0 1.0\nc[8] 1.49 0.02 0.31 0.91 1.25 1.49 1.69 2.15 286.0 1.0\n\nSamples were drawn using NUTS(diag_e) at Sat Jun 14 13:13:03 2014.\nFor each parameter, n_eff is a crude measure of effective sample size,\nand Rhat is the potential scale reduction factor on split chains (at \nconvergence, Rhat=1).\n" ], [ "w=loadStanFit('fit4')\npt=w['pt']\npc=w['pc']\nDP=np.zeros((pt.shape[2]+2,pt.shape[1],2))\nDP[0,:,:]=1\nDP[1:-1,:,:]=np.array([pc,pt]).T.mean(2)\nDP=-np.diff(DP,axis=0)\nplotComparison(DS[:,:7],DP,stan=True)", "_____no_output_____" ], [ "pystanErrorbar(w,keys=['beta','c','delt'])", "delt 0.673595459215 CI [0.211,1.140]\n" ], [ "dat = {'y':np.int32(DR[:,1:7])+1,'x':np.int32(DR[:,0]),'N':DR.shape[0] ,'K':10,'M':6}\nfit5 = sm4.sampling(data=dat,iter=1000, chains=4,thin=2,warmup=500,njobs=2,seed=4)\nprint pystan.misc._print_stanfit(fit4,pars=['delt','bbeta','c'],digits_summary=2)\nsaveStanFit(fit5,'fit5')", "INFO:pystan:NOW ON CHAIN 1\nINFO:pystan:NOW ON CHAIN 0\nINFO:pystan:NOW ON CHAIN 3\nINFO:pystan:NOW ON CHAIN 2\n" ], [ "w=loadStanFit('fit5')\npt=w['pt']\npc=w['pc']\nDP=np.zeros((pt.shape[2]+2,pt.shape[1],2))\nDP[0,:,:]=1\nDP[1:-1,:,:]=np.array([pc,pt]).T.mean(2)\nDP=-np.diff(DP,axis=0)\nplotComparison(DR[:,:7],DP,stan=True)", "_____no_output_____" ], [ "pystanErrorbar(w,keys=['beta','c','delt'])", "delt -0.0163305817326 CI [-0.220,0.186]\n" ], [ "model='''\ndata {\n int<lower=2> K;\n int<lower=0> N;\n int<lower=1> M;\n int<lower=1,upper=K> y[N,M];\n int x[N,2];\n}\nparameters {\n // real mb; real<lower=0,upper=100> sb[2];\n vector[M-1] bbeta;\n real dd[3];\n ordered[K-1] c;\n}\ntransformed parameters{\n //real pt[M,K-1]; real pc[M,K-1];\n vector[M] beta;\n for (m in 1:M){\n if (m==1) beta[m]<-0.0; \n else beta[m]<-bbeta[m-1];\n //for (k in 1:(K-1)){\n // pt[m,k] <- inv_logit(beta[m]+delt-c[k]);\n // pc[m,k] <- inv_logit(beta[m]-c[k]);}\n}}\nmodel {\nfor (k in 1:(K-1)) c[k]~ uniform(-100,100);\nfor (m in 1:M){\n for (n in 1:N) y[n,m] ~ ordered_logistic(beta[m]\n +dd[2]*x[n,1]*(1-x[n,2]) // rep + control\n +dd[1]*x[n,2]*(1-x[n,1]) // orig + treat\n +dd[3]*x[n,1]*x[n,2], c); // rep + treat\n}}\n'''\nsm5=pystan.StanModel(model_code=model)", "INFO:pystan:COMPILING THE C++ CODE FOR MODEL anon_model_5463f0c280d4ea208ffd631f059de31d NOW.\n" ], [ "dat = {'y':np.int32(D[:,2:8])+1,'x':np.int32(D[:,[0,1]]),'N':D.shape[0] ,'K':10,'M':6}\nfit6 = sm5.sampling(data=dat,iter=1000, chains=4,thin=2,warmup=500,njobs=2,seed=4)\nprint pystan.misc._print_stanfit(fit6,pars=['dd','bbeta','c'],digits_summary=2)\nsaveStanFit(fit6,'fit6')", "Inference for Stan model: anon_model_5463f0c280d4ea208ffd631f059de31d.\n4 chains, each with iter=1000; warmup=500; thin=2; \npost-warmup draws per chain=250, total post-warmup draws=1000.\n\n mean se_mean sd 2.5% 25% 50% 75% 97.5% n_eff Rhat\ndd[0] 0.67 0.01 0.22 0.22 0.52 0.67 0.82 1.12 307.0 1.0\ndd[1] 1.23 0.01 0.18 0.88 1.11 1.24 1.35 1.58 307.0 1.0\ndd[2] 1.21 0.01 0.18 0.86 1.09 1.21 1.33 1.57 306.0 1.0\nbbeta[0] -3.4 0.01 0.18-3.75-3.53-3.39-3.27 -3.06 317.0 1.0\nbbeta[1]-0.44 0.01 0.16-0.75-0.56-0.43-0.33 -0.11 295.0 1.0\nbbeta[2]-0.24 0.01 0.17-0.59-0.35-0.25-0.12 0.09 309.0 1.0\nbbeta[3]-0.61 0.01 0.17-0.94-0.73-0.62-0.49 -0.28 296.0 1.0\nbbeta[4] 0.57 0.01 0.18 0.23 0.45 0.56 0.69 0.92 293.0 1.01\nc[0] -4.42 0.02 0.27-4.94-4.61-4.43-4.23 -3.94 311.0 1.0\nc[1] -3.41 0.01 0.24-3.89-3.58-3.41-3.25 -2.95 314.0 1.0\nc[2] -2.52 0.01 0.22-2.95-2.67-2.52-2.37 -2.08 303.0 1.0\nc[3] -1.21 0.01 0.21 -1.6-1.35 -1.2-1.07 -0.79 298.0 1.0\nc[4] -0.82 0.01 0.2-1.19-0.96-0.82-0.68 -0.4 294.0 1.01\nc[5] -0.3 0.01 0.2-0.69-0.45-0.31-0.17 0.09 295.0 1.0\nc[6] 0.38 0.01 0.2-0.01 0.24 0.38 0.51 0.8 294.0 1.01\nc[7] 0.79 0.01 0.2 0.4 0.65 0.79 0.92 1.2 296.0 1.0\nc[8] 1.26 0.01 0.2 0.87 1.12 1.26 1.39 1.68 295.0 1.01\n\nSamples were drawn using NUTS(diag_e) at Sat Jun 14 13:46:31 2014.\nFor each parameter, n_eff is a crude measure of effective sample size,\nand Rhat is the potential scale reduction factor on split chains (at \nconvergence, Rhat=1).\n" ], [ "w=loadStanFit('fit6')\npystanErrorbar(w,keys=['beta','c','dd'])", "_____no_output_____" ], [ "plt.figure(figsize=(10,4))\nc=w['c']\nb=w['beta']\nd=w['dd']\nerrorbar(c,x=np.linspace(6.5,8,9))\nax=plt.gca()\nplt.plot([-1,100],[0,0],'k',lw=2)\nax.set_yticks(np.median(c,axis=0))\nax.set_yticklabels(np.arange(1,10)+0.5)\nplt.grid(b=False,axis='x')\nerrorbar(b[:,::-1],x=np.arange(9,15),clr='g')\nerrorbar(d,x=np.arange(15,18),clr='r')\nplt.xlim([6,17.5])\nax.set_xticks(range(9,18))\nax.set_xticklabels(il[:6][::-1]+['OT','RC','RT'])\nfor i in range(d.shape[1]): printCI(d[:,i])\nprintCI(d[:,2]-d[:,1])", "var 0.670, CI 0.225, 1.119\nvar 1.233, CI 0.882, 1.575\nvar 1.213, CI 0.874, 1.566\nvar -0.020, CI -0.212, 0.180\n" ], [ "\nc", "_____no_output_____" ], [ "def ordinalLogitRvs(beta, c,n,size=1):\n assert np.all(np.diff(c)>0) # c must be strictly increasing\n def invLogit(x): return 1/(1+np.exp(-x))\n p=[1]+list(invLogit(beta-c))+[0]\n p=-np.diff(p)\n #return np.random.multinomial(n,p,size)\n return np.int32(np.round(p*n))\ndef reformatData(dat):\n out=[]\n for k in range(dat.size):\n out.extend([k]*dat[k])\n return np.array(out)\nb=np.linspace(-10,7,21)\nd=np.median(w['dd'][:,0])\nc=np.median(w['c'],axis=0)\nS=[];P=[]\nfor bb in b:\n S.append([np.squeeze(ordinalLogitRvs(bb,c,100)),\n np.squeeze(ordinalLogitRvs(bb+d,c,100))])\n P.append([reformatData(S[-1][0]),reformatData(S[-1][1])])", "_____no_output_____" ], [ "model='''\ndata {\n int<lower=2> K;\n int<lower=0> y1[K];\n int<lower=0> y2[K];\n}\nparameters {\n real<lower=-1000,upper=1000> d;\n ordered[K-1] c;\n}\nmodel {\nfor (k in 1:(K-1)) c[k]~ uniform(-200,200);\nfor (k in 1:K){\n for (n in 1:y1[k]) k~ ordered_logistic(0.0,c);\n for (n in 1:y2[k]) k~ ordered_logistic(d ,c);\n}}\n'''\nsm9=pystan.StanModel(model_code=model)", "_____no_output_____" ], [ "#(S[k][0]!=0).sum()+1\nfor k in range(21): \n i1=np.nonzero(S[k][0]!=0)[0]\n i2=np.nonzero(S[k][1]!=0)[0]\n if max((S[k][0]!=0).sum(),(S[k][1]!=0).sum())<9:\n s= max(min(i1[0],i2[0])-1,0)\n e= min(max(i1[-1],i2[-1])+1,10)\n S[k][0]=S[k][0][s:e+1]\n S[k][1]=S[k][1][s:e+1]", "_____no_output_____" ], [ "S[0][0].size", "_____no_output_____" ], [ "ds=[];cs=[]\nfor k in range(len(S)):\n dat = {'y1':S[k][0],'y2':S[k][1],'K':S[k][0].size}\n fit = sm9.sampling(data=dat,iter=1000, chains=4,thin=2,warmup=500,njobs=2,seed=4)\n print fit\n saveStanFit(fit,'dc%d'%k)", "Inference for Stan model: anon_model_cdc057088a7ff9c9329245d2161c83d6.\n4 chains, each with iter=1000; warmup=500; thin=2; \npost-warmup draws per chain=250, total post-warmup draws=1000.\n\n mean se_mean sd 2.5% 25% 50% 75% 97.5% n_eff Rhat\nd -451.0 25.8311.4-970.4-714.4-435.7-200.2 77.5 146.0 1.0\nc[0] 99.7 4.8 56.2 9.7 52.6100.1146.0 194.5 138.0 1.0\nlp__ 5.7 0.1 0.7 3.4 5.5 6.0 6.2 6.2 147.0 1.0\n\nSamples were drawn using NUTS(diag_e) at Sat Jun 14 21:39:48 2014.\nFor each parameter, n_eff is a crude measure of effective sample size,\nand Rhat is the potential scale reduction factor on split chains (at \nconvergence, Rhat=1).\n" ], [ "for k in range(21): \n i1=np.nonzero(S[k][0]!=0)[0]\n i2=np.nonzero(S[k][1]!=0)[0]\n if max((S[k][0]!=0).sum(),(S[k][1]!=0).sum())<9:\n s= min(i1[0],i2[0])\n e= max(i1[-1],i2[-1])\n S[k][0]=S[k][0][s:e+1]\n S[k][1]=S[k][1][s:e+1]", "_____no_output_____" ], [ "ds=[];cs=[]\nfor k in range(len(S)):\n if S[k][0].size==1: continue\n dat = {'y1':S[k][0],'y2':S[k][1],'K':S[k][0].size}\n fit = sm9.sampling(data=dat,iter=1000, chains=4,thin=2,warmup=500,njobs=2,seed=4)\n #print fit\n saveStanFit(fit,'dd%d'%k)", "Inference for Stan model: anon_model_cdc057088a7ff9c9329245d2161c83d6.\n4 chains, each with iter=1000; warmup=500; thin=2; \npost-warmup draws per chain=250, total post-warmup draws=1000.\n\n mean se_mean sd 2.5% 25% 50% 75% 97.5% n_eff Rhat\nd 0.0 0.1 1.8 -3.7 -1.0 0.1 1.2 3.4 197.0 1.0\nc[0] 5.2 0.1 1.2 3.3 4.3 5.0 6.0 8.0 181.0 1.0\nlp__ -6.1 0.1 1.1 -9.1 -6.6 -5.8 -5.3 -5.0 221.0 1.0\n\nSamples were drawn using NUTS(diag_e) at Sat Jun 14 23:06:03 2014.\nFor each parameter, n_eff is a crude measure of effective sample size,\nand Rhat is the potential scale reduction factor on split chains (at \nconvergence, Rhat=1).\n" ], [ "ds=[];xs=[]\nfor k in range(b.size):\n try:\n f=loadStanFit('dd%d'%k)['d']\n xs.append(b[k])\n ds.append(f)\n except:pass\nds=np.array(ds);xs=np.array(xs)\nds.shape", "_____no_output_____" ], [ "d1=np.median(w['dd'][:,0])\nd2=DS[DS[:,0]==1,-2].mean()-DS[DS[:,0]==0,-2].mean()\nplt.figure(figsize=(8,4))\nplt.plot([-10,5],[d1,d1],'r',alpha=0.5)\nres1=errorbar(ds.T,x=xs-0.1)\nax1=plt.gca()\nplt.ylim([-2,2])\nplt.xlim([-10,5])\nplt.grid(b=False,axis='x')\nax2 = ax1.twinx()\nres2=np.zeros((b.size,5))\nfor k in range(b.size):\n res2[k,:]=plotCIttest2(y1=P[k][0],y2=P[k][1],x=b[k]+0.1)\nplt.ylim([-2/d1*d2,2/d1*d2])\nplt.xlim([-10,5])\nplt.grid(b=False,axis='y')\nplt.plot(np.median(w['beta'],axis=0),[-0.9]*6,'ob')\nplt.plot(np.median(w['beta']+np.atleast_2d(w['dd'][:,1]).T,axis=0),[-1.1]*6,'og')", "_____no_output_____" ], [ "d1=np.median(w['dd'][:,0])\nd2=DS[DS[:,0]==1,-2].mean()-DS[DS[:,0]==0,-2].mean()\nplt.figure(figsize=(8,4))\nax1=plt.gca()\nplt.plot([-10,5],[d1,d1],'r',alpha=0.5)\ntemp=[list(xs)+list(xs)[::-1],list(res1[:,1])+list(res1[:,2])[::-1]]\nax1.add_patch(plt.Polygon(xy=np.array(temp).T,alpha=0.2,fc='k',ec='k'))\nplt.plot(xs,res1[:,0],'k')\nplt.ylim([-1.5,2])\nplt.xlim([-10,5])\nplt.grid(b=False,axis='x')\nplt.legend(['True ES','Estimate Ordinal Logit'],loc=8)\nplt.ylabel('Estimate Ordinal Logit')\nax2 = ax1.twinx()\ntemp=[list(b)+list(b)[::-1],list(res2[:,1])+list(res2[:,2])[::-1]]\nfor t in range(len(temp[0]))[::-1]:\n if np.isnan(temp[1][t]):\n temp[0].pop(t);temp[1].pop(t)\nax2.add_patch(plt.Polygon(xy=np.array(temp).T,alpha=0.2,fc='m',ec='m'))\nplt.plot(b,res2[:,0],'m')\nplt.ylim([-1.5/d1*d2,2/d1*d2])\nplt.xlim([-10,5])\nplt.grid(b=False,axis='y')\nplt.plot(np.median(w['beta'],axis=0),[-0.3]*6,'ob')\nplt.plot(np.median(w['beta']+np.atleast_2d(w['dd'][:,1]).T,axis=0),[-0.5]*6,'og')\nplt.legend(['Estimate T-C','Item Difficulty Orignal Study','Item Difficulty Replication'],loc=4)\nplt.ylabel('Estimate T - C',color='m')\nfor tl in ax2.get_yticklabels():tl.set_color('m')", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/chiranjeet14/ML_Journey/blob/master/Hackerearth-Predict_condition_and_insurance_amount/train_models.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ] ], [ [ "!pip3 install xgboost > /dev/null", "_____no_output_____" ], [ "import pandas as pd\nimport numpy as np\nimport io\nimport gc\nimport time\nfrom pprint import pprint\n# import PIL.Image as Image\n# import matplotlib.pylab as plt\nfrom datetime import date\n\n# import tensorflow as tf\n# import tensorflow_hub as hub\n\n# settings\nimport warnings\nwarnings.filterwarnings(\"ignore\")\ngc.enable()", "_____no_output_____" ], [ "# Calculating Precision, Recall and f1-score\ndef model_score(actual_value,predicted_values):\n from sklearn.metrics import confusion_matrix \n from sklearn.metrics import accuracy_score \n from sklearn.metrics import classification_report \n from sklearn.metrics import recall_score\n \n actual = actual_value\n predicted = predicted_values\n results = confusion_matrix(actual, predicted) \n \n print('Confusion Matrix :')\n print(results) \n print('Accuracy Score :',accuracy_score(actual, predicted))\n print('Report : ')\n print(classification_report(actual, predicted))\n print('Recall Score : ')\n print(recall_score(actual, predicted))", "_____no_output_____" ], [ "# connect to google drive\nfrom google.colab import drive\ndrive.mount('/content/drive')", "Mounted at /content/drive\n" ], [ "gDrivePath = '/content/drive/MyDrive/Datasets/Hackerearth_vehicle_insurance_claim/dataset/'\ngDriveTrainFinal = gDrivePath + 'final_datasets/train_final.csv'\ngDriveTestFinal = gDrivePath + 'final_datasets/test_final.csv'", "_____no_output_____" ], [ "df_train = pd.read_csv(gDriveTrainFinal)\ndf_test = pd.read_csv(gDriveTestFinal)", "_____no_output_____" ], [ "df_train.head()", "_____no_output_____" ], [ "df_train.sample(n = 10)", "_____no_output_____" ], [ "df_train.drop(['image_name'], axis=1, inplace=True)\ndf_test.drop(['image_name'], axis=1, inplace=True)", "_____no_output_____" ], [ "df_train[['Insurance_company', 'Cost_of_vehicle', 'Min_coverage', 'Max_coverage', 'Condition', 'Amount']].isna().any()", "_____no_output_____" ] ], [ [ "### Removing NaN in target variable", "_____no_output_____" ] ], [ [ "# select rows where amount is not NaN\ndf_train = df_train[df_train['Amount'].notna()]\ndf_train[df_train['Amount'].isna()].shape", "_____no_output_____" ], [ "# delete rows where Amount < 0\ndf_train = df_train[df_train['Amount'] >= 0]\ndf_train[['Cost_of_vehicle', 'Min_coverage', 'Max_coverage', 'Amount']].describe()", "_____no_output_____" ], [ "selected_columns = ['Cost_of_vehicle', 'Min_coverage', 'Max_coverage']\n\n# replacing nan values with median\nfrom sklearn.impute import SimpleImputer \nimputer = SimpleImputer(missing_values = np.nan, strategy ='median') \nimputer = imputer.fit(df_train[selected_columns]) \n\n# Imputing the data \ndf_train[selected_columns] = imputer.transform(df_train[selected_columns])\ndf_test[selected_columns] = imputer.transform(df_test[selected_columns])", "_____no_output_____" ], [ "df_train[['Insurance_company', 'Cost_of_vehicle', 'Min_coverage', 'Max_coverage', 'Condition', 'Amount']].isna().any()", "_____no_output_____" ] ], [ [ "### Checking if the dataset is balanced/imbalanced - Condition", "_____no_output_____" ] ], [ [ "# python check if dataset is imbalanced : https://www.kaggle.com/rafjaa/resampling-strategies-for-imbalanced-datasets\n\ntarget_count = df_train['Condition'].value_counts()\nprint('Class 0 (No):', target_count[0])\nprint('Class 1 (Yes):', target_count[1])\nprint('Proportion:', round(target_count[0] / target_count[1], 2), ': 1')\n\ntarget_count.plot(kind='bar', title='Condition')", "Class 0 (No): 99\nClass 1 (Yes): 1288\nProportion: 0.08 : 1\n" ] ], [ [ "### Splitting Data into train-cv", "_____no_output_____" ] ], [ [ "classification_labels = df_train['Condition'].values\n\n# for regresion delete rows where Condition = 0\ndf_train_regression = df_train[df_train['Condition'] == 1]\nregression_labels = df_train_regression['Amount'].values\n######\n\ndf_train_regression.drop(['Condition','Amount'], axis=1, inplace=True)\n\ndf_train.drop(['Condition','Amount'], axis=1, inplace=True)\ndf_test.drop(['Condition','Amount'], axis=1, inplace=True, errors='ignore')", "_____no_output_____" ], [ "# classification split\nfrom sklearn.model_selection import train_test_split\nX_train, X_cv, y_train, y_cv = train_test_split(df_train, classification_labels, test_size=0.1)", "_____no_output_____" ] ], [ [ "### Over Sampling using SMOTE\n", "_____no_output_____" ] ], [ [ "# https://machinelearningmastery.com/smote-oversampling-for-imbalanced-classification/\nfrom imblearn.over_sampling import SMOTE\nsmote_overSampling = SMOTE()\nX_train,y_train = smote_overSampling.fit_resample(X_train,y_train)\nunique, counts = np.unique(y_train, return_counts=True)\ndict(zip(unique, counts))", "_____no_output_____" ] ], [ [ "### Scaling data", "_____no_output_____" ] ], [ [ "from sklearn.preprocessing import StandardScaler\nscaler = StandardScaler()\nX_train_scaled = scaler.fit_transform(X_train)\nX_cv_scaled = scaler.transform(X_cv)\n\nX_test_scaled = scaler.transform(df_test)\nX_train_scaled", "_____no_output_____" ] ], [ [ "## Modelling & Cross-Validation", "_____no_output_____" ], [ "### Classification", "_____no_output_____" ] ], [ [ "%%time\n# Train multiple models : https://www.kaggle.com/tflare/testing-multiple-models-with-scikit-learn-0-79425\nfrom sklearn.linear_model import LogisticRegression\nfrom sklearn.svm import SVC, LinearSVC\nfrom sklearn.neighbors import KNeighborsClassifier\nfrom sklearn.tree import DecisionTreeClassifier\nfrom sklearn.ensemble import RandomForestClassifier\nfrom sklearn.ensemble import AdaBoostClassifier\nfrom sklearn.ensemble import BaggingClassifier\nfrom sklearn.ensemble import ExtraTreesClassifier\nfrom sklearn.ensemble import GradientBoostingClassifier\nfrom sklearn.linear_model import LogisticRegressionCV\nfrom xgboost import XGBClassifier\n\nfrom sklearn.model_selection import cross_val_score\n\nmodels = []\n\nLogisticRegression = LogisticRegression(n_jobs=-1)\nLinearSVC = LinearSVC()\nKNeighbors = KNeighborsClassifier(n_jobs=-1)\nDecisionTree = DecisionTreeClassifier()\nRandomForest = RandomForestClassifier()\nAdaBoost = AdaBoostClassifier()\nBagging = BaggingClassifier()\nExtraTrees = ExtraTreesClassifier()\nGradientBoosting = GradientBoostingClassifier()\nLogisticRegressionCV = LogisticRegressionCV(n_jobs=-1)\nXGBClassifier = XGBClassifier(nthread=-1)\n\n# models.append((\"LogisticRegression\",LogisticRegression))\n# models.append((\"LinearSVC\", LinearSVC))\n# models.append((\"KNeighbors\", KNeighbors))\n# models.append((\"DecisionTree\", DecisionTree))\n# models.append((\"RandomForest\", RandomForest))\nmodels.append((\"AdaBoost\", AdaBoost))\n# models.append((\"Bagging\", Bagging))\n# models.append((\"ExtraTrees\", ExtraTrees))\n# models.append((\"GradientBoosting\", GradientBoosting))\n# models.append((\"LogisticRegressionCV\", LogisticRegressionCV))\n# models.append((\"XGBClassifier\", XGBClassifier))\n\n# metric_names = ['f1', 'average_precision', 'accuracy', 'precision', 'recall']\nmetric_names = ['f1']\nresults = []\nnames = []\n\nnested_dict = {}\n\nfor name,model in models:\n nested_dict[name] = {}\n for metric in metric_names:\n print(\"\\nRunning : {}, with metric : {}\".format(name, metric))\n score = cross_val_score(model, X_train_scaled, y_train, n_jobs=-1, scoring=metric, cv=5)\n nested_dict[name][metric] = score.mean()", "\nRunning : AdaBoost, with metric : f1\nCPU times: user 399 ms, sys: 150 ms, total: 548 ms\nWall time: 1min 13s\n" ], [ "import json\nprint(json.dumps(nested_dict, sort_keys=True, indent=4))", "{\n \"AdaBoost\": {\n \"f1\": 0.9991397849462367\n }\n}\n" ] ], [ [ "### Regression", "_____no_output_____" ] ], [ [ "X_train_regression, X_cv_regression, y_train_regression, y_cv_regression = train_test_split(df_train_regression, regression_labels, test_size=0.1)\n\nscaler = StandardScaler()\nX_train_scaled_regression = scaler.fit_transform(X_train_regression)\nX_cv_scaled_regression = scaler.transform(X_cv_regression)\n\nX_test_scaled_regression = scaler.transform(df_test)\nX_train_scaled_regression", "_____no_output_____" ], [ "%%time\nfrom sklearn.linear_model import LinearRegression, SGDRegressor\nfrom sklearn.svm import SVR, LinearSVR\nfrom sklearn.neighbors import KNeighborsRegressor\nfrom sklearn.tree import DecisionTreeRegressor\nfrom sklearn.ensemble import RandomForestRegressor\nfrom sklearn.ensemble import AdaBoostRegressor\nfrom sklearn.ensemble import BaggingRegressor\nfrom sklearn.ensemble import ExtraTreesRegressor\nfrom sklearn.ensemble import GradientBoostingRegressor\nfrom xgboost import XGBClassifier\n\nfrom sklearn.model_selection import cross_val_score\n\nmodels = []\n\nLinearReg = LinearRegression(n_jobs=-1)\nSGDReg = SGDRegressor()\nSVReg = SVR()\nLinearSVReg = LinearSVR()\nKNeighborsReg = KNeighborsRegressor(n_jobs=-1)\nDecisionTreeReg = DecisionTreeRegressor()\nRandomForestReg = RandomForestRegressor(n_jobs=-1)\nAdaBoostReg = AdaBoostRegressor()\nBaggingReg = BaggingRegressor(n_jobs=-1)\nExtraTreesReg = ExtraTreesRegressor(n_jobs=-1)\nGradientBoostingReg = GradientBoostingRegressor()\n# XGBReg = XGBRegressor(nthread=-1)\n\n# models.append((\"LinearRegression\",LinearReg))\n# models.append((\"SGDRegressor\",SGDReg))\n# models.append((\"SVR\", SVReg))\n# models.append((\"LinearSVR\", LinearSVReg))\n# models.append((\"KNeighborsRegressor\", KNeighborsReg))\n# models.append((\"DecisionTreeRegressor\", DecisionTreeReg))\n# models.append((\"RandomForestRegressor\", RandomForestReg))\n# models.append((\"AdaBoostRegressor\", AdaBoostReg))\n# models.append((\"BaggingRegressor\", BaggingReg))\nmodels.append((\"ExtraTreesRegressor\", ExtraTreesReg))\n# models.append((\"GradientBoostingRegressor\", GradientBoostingReg))\n# models.append((\"XGBReg\", XGBRegressor))\n\n# metric_names = ['f1', 'average_precision', 'accuracy', 'precision', 'recall']\nmetric_names = ['r2']\nresults = []\nnames = []\n\nnested_dict = {}\n\n# for name,model in models:\n# nested_dict[name] = {}\n# for metric in metric_names:\n# print(\"\\nRunning : {}, with metric : {}\".format(name, metric))\n# score = cross_val_score(model, X_train_scaled_regression, y_train_regression, n_jobs=-1, scoring=metric, cv=5)\n# nested_dict[name][metric] = score.mean()", "CPU times: user 426 µs, sys: 0 ns, total: 426 µs\nWall time: 434 µs\n" ], [ "# import json\n# print(json.dumps(nested_dict, sort_keys=True, indent=4))", "_____no_output_____" ], [ "# # Hyperparameter tuning ExtraTreesRegressor\n\n# # ExtraTreesRegressor(bootstrap=True, criterion='mae',n_estimators=100, warm_start=True,\n\n# # max_depth=None, max_features='auto', max_leaf_nodes=None,\n# # max_samples=None, min_impurity_decrease=0.0,\n# # min_impurity_split=None, min_samples_leaf=1,\n# # min_samples_split=2, min_weight_fraction_leaf=0.0,\n# # n_jobs=-1, oob_score=False,\n# # random_state=None, verbose=0)\n\nfrom sklearn.model_selection import GridSearchCV, RandomizedSearchCV\n\nmodel = ExtraTreesRegressor(n_jobs=-1, bootstrap=True, criterion='mae', warm_start=True, max_depth=9, max_features='auto')\n\nparam_grid = {\n # 'n_estimators': np.arange(100, 3000, 100, dtype=int),\n # 'criterion': ['mse', 'mae'],\n # 'max_depth': np.arange(5, 16, 1, dtype=int),\n # 'bootstrap': [True, False],\n # 'max_features': ['auto', 'sqrt', 'log2'],\n # 'max_features': np.arange(100, 1540, 20, dtype=int),\n # 'warm_start': [True, False],\n}\n \ngsc = GridSearchCV(estimator=model, param_grid=param_grid, scoring='r2', cv=5, n_jobs=-1, verbose=1000)\ngrid_result = gsc.fit(X_train_scaled_regression, y_train_regression)\n\n# n_iter_search = 100\n# random_search = RandomizedSearchCV(estimator=model, param_distributions=param_grid, n_iter=n_iter_search, scoring='r2', cv=3, n_jobs=-1, verbose=500)\n# random_search.fit(X_train, y_train)", "Fitting 5 folds for each of 1 candidates, totalling 5 fits\n[Parallel(n_jobs=-1)]: Using backend LokyBackend with 2 concurrent workers.\n[Parallel(n_jobs=-1)]: Done 1 tasks | elapsed: 8.9min\n[Parallel(n_jobs=-1)]: Done 2 tasks | elapsed: 9.0min\n[Parallel(n_jobs=-1)]: Done 3 out of 5 | elapsed: 17.5min remaining: 11.7min\n[Parallel(n_jobs=-1)]: Done 5 out of 5 | elapsed: 22.3min remaining: 0.0s\n[Parallel(n_jobs=-1)]: Done 5 out of 5 | elapsed: 22.3min finished\n" ], [ "print(\"Best: %f using %s\" % (grid_result.best_score_, grid_result.best_params_))", "Best: 0.062463 using {}\n" ] ], [ [ "### Predicting on CV data", "_____no_output_____" ] ], [ [ "classification_alg = AdaBoost\n# regression_alg = ExtraTreesReg\n\n# hypertuned model\nregression_alg = gsc\n\nclassification_alg.fit(X_train_scaled, y_train)\nregression_alg.fit(X_train_scaled_regression, y_train_regression)\n\n\n# predictions_class = classification_alg.predict(X_cv)\n# pprint(classification_alg.get_params())", "Fitting 5 folds for each of 1 candidates, totalling 5 fits\n[Parallel(n_jobs=-1)]: Using backend LokyBackend with 2 concurrent workers.\n[Parallel(n_jobs=-1)]: Done 1 tasks | elapsed: 9.0min\n[Parallel(n_jobs=-1)]: Done 2 tasks | elapsed: 9.1min\n[Parallel(n_jobs=-1)]: Done 3 out of 5 | elapsed: 17.5min remaining: 11.7min\n[Parallel(n_jobs=-1)]: Done 5 out of 5 | elapsed: 22.4min remaining: 0.0s\n[Parallel(n_jobs=-1)]: Done 5 out of 5 | elapsed: 22.4min finished\n" ], [ "# model_score(y_cv,predictions)", "_____no_output_____" ] ], [ [ "### Predicting on test Data", "_____no_output_____" ] ], [ [ "trained_classifier = classification_alg\ntrained_regressor = regression_alg\n\npredictions_trained_classifier_test = trained_classifier.predict(X_test_scaled)\npredictions_trained_regressor_test = trained_regressor.predict(X_test_scaled_regression)", "_____no_output_____" ], [ "read = pd.read_csv(gDrivePath + 'test.csv')\nsubmission = pd.DataFrame({\n \"Image_path\": read[\"Image_path\"],\n \"Condition\": predictions_trained_classifier_test,\n \"Amount\": predictions_trained_regressor_test\n })", "_____no_output_____" ], [ "submission.head()", "_____no_output_____" ], [ "submission['Amount'][submission.Condition == 0] = 0", "_____no_output_____" ], [ "submission[submission['Condition'] == 0].sample(n = 10)", "_____no_output_____" ], [ "submission.Amount = submission.Amount.round()\nsubmission.head()", "_____no_output_____" ], [ "submission.to_csv('./submission.csv', index=False)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Build a Traffic Sign Recognition Classifier Deep Learning", "_____no_output_____" ], [ "Some improvements are taken :\n- [x] Adding of convolution networks at the same size of previous layer, to get 1x1 layer\n- [x] Activation function use 'ReLU' instead of 'tanh'\n- [x] Adaptative learning rate, so learning rate is decayed along to training phase\n- [x] Enhanced training dataset", "_____no_output_____" ], [ "## Load and Visualize the Enhanced training dataset\n\nFrom the original standard German Traffic Signs dataset, we add some 'generalized' sign to cover cases that the classifier can not well interpret small figures inside. \n`Also`, in our Enhanced training dataset, each figure is taken from standard library - not from road images, so they are very clear and in high-definition.", "_____no_output_____" ], [ "*Enhanced traffic signs &#8595;* \n<img src=\"figures/enhanced_training_dataset_text.png\" alt=\"Drawing\" style=\"width: 600px;\"/>", "_____no_output_____" ] ], [ [ "# load enhanced traffic signs\nimport os\nimport cv2\nimport matplotlib.pyplot as plot\nimport numpy\ndir_enhancedsign = 'figures\\enhanced_training_dataset2'\nfiles_enhancedsign = [os.path.join(dir_enhancedsign, f) for f in os.listdir(dir_enhancedsign)]\n\n# read & resize (32,32) images in enhanced dataset\nimages_enhancedsign = numpy.array([cv2.cvtColor(cv2.resize(cv2.imread(f), (32,32), interpolation = cv2.INTER_AREA), cv2.COLOR_BGR2RGB) for f in files_enhancedsign])\n\n# plot new test images\nfig, axes = plot.subplots(7, 8)\nplot.suptitle('Enhanced training dataset')\nfor i, ax in enumerate(axes.ravel()):\n if i < 50:\n ax.imshow(images_enhancedsign[i])\n# ax.set_title('{}'.format(i))\n plot.setp(ax.get_xticklabels(), visible=False)\n plot.setp(ax.get_yticklabels(), visible=False)\n ax.set_xticks([]), ax.set_yticks([])\n \n ax.axis('off')\nplot.draw()\nfig.savefig('figures/' + 'enhancedsign' + '.jpg', dpi=700)\n\nprint(\"Image Shape : {}\".format(images_enhancedsign[0].shape))\nprint()\nprint(\"Enhanced Training Dataset : {} samples\".format(len(images_enhancedsign)))", "Image Shape : (32, 32, 3)\n\nEnhanced Training Dataset : 50 samples\n" ], [ "# classes of enhanced dataset are taken from their filenames\nimport re\nregex = re.compile(r'\\d+')\n\ny_enhancedsign = [int(regex.findall(f)[0]) for f in os.listdir(dir_enhancedsign)]\nprint(y_enhancedsign)", "[0, 1, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 2, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 3, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 4, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 5, 6, 7, 8, 9]\n" ] ], [ [ "*Enhanced German traffic signs dataset &#8595;* \n<img src=\"figures/enhanced_training_dataset.png\" alt=\"Drawing\" style=\"width: 600px;\"/>", "_____no_output_____" ], [ "**We would have 50 classes in total with new enhanced training dataset :**", "_____no_output_____" ] ], [ [ "n_classes_enhanced = len(numpy.unique(y_enhancedsign))\n\nprint('n_classes enhanced : {}'.format(n_classes_enhanced))", "n_classes enhanced : 50\n" ] ], [ [ "## Load and Visualize the standard German Traffic Signs Dataset", "_____no_output_____" ] ], [ [ "# Load pickled data\nimport pickle\nimport numpy\n\n# TODO: Fill this in based on where you saved the training and testing data\n\ntraining_file = 'traffic-signs-data/train.p'\nvalidation_file = 'traffic-signs-data/valid.p'\ntesting_file = 'traffic-signs-data/test.p'\n\nwith open(training_file, mode='rb') as f:\n train = pickle.load(f)\nwith open(validation_file, mode='rb') as f:\n valid = pickle.load(f)\nwith open(testing_file, mode='rb') as f:\n test = pickle.load(f)\n \nX_train, y_train = train['features'], train['labels'] # training dataset\nX_valid, y_valid = valid['features'], valid['labels'] # validation dataset used in training phase\nX_test, y_test = test['features'], test['labels'] # test dataset\nn_classes_standard = len(numpy.unique(y_train))\n\nassert(len(X_train) == len(y_train))\nassert(len(X_valid) == len(y_valid))\nassert(len(X_test) == len(y_test))\nprint(\"Image Shape : {}\".format(X_train[0].shape))\nprint()\nprint(\"Training Set : {} samples\".format(len(X_train)))\nprint(\"Validation Set : {} samples\".format(len(X_valid)))\nprint(\"Test Set : {} samples\".format(len(X_test)))\nprint('n_classes standard : {}'.format(n_classes_standard))", "Image Shape : (32, 32, 3)\n\nTraining Set : 34799 samples\nValidation Set : 4410 samples\nTest Set : 12630 samples\nn_classes standard : 43\n" ], [ "n_classes = n_classes_enhanced", "_____no_output_____" ] ], [ [ "## Implementation of LeNet\n\n>http://yann.lecun.com/exdb/publis/pdf/sermanet-ijcnn-11.pdf \nAbove is the original article of Pierre Sermanet and Yann LeCun in 1998 that we can follow to create LeNet convolutional networks with a good accuracy even for very-beginners in deep-learning. \nIt's really excited to see that many years of works now could be implemented in just 9 lines of code thank to Keras high-level API !\n(low-level API implementation with Tensorflow 1 is roughly 20 lines of code) \n\n>Here is also an interesting medium article : \nhttps://medium.com/@mgazar/lenet-5-in-9-lines-of-code-using-keras-ac99294c8086 ", "_____no_output_____" ] ], [ [ "### Import tensorflow and keras\nimport tensorflow as tf\nfrom tensorflow import keras\n\nprint (\"TensorFlow version: \" + tf.__version__)", "TensorFlow version: 2.1.0\n" ] ], [ [ "### 2-stage ConvNet architecture by Pierre Sermanet and Yann LeCun\n\nWe will try to implement the 2-stage ConvNet architecture by Pierre Sermanet and Yann LeCun which is not sequential. \n\nKeras disposes keras.Sequential() API for sequential architectures but it can not handle models with non-linear topology, shared layers or multi-in/output. So the choose of the 2-stage ConvNet architecture by `Pierre Sermanet` and `Yann LeCun` is to challenge us also.\n\n<img src=\"figures/lenet_2.png\" alt=\"Drawing\" style=\"width: 550px;\"/>\n\n>Source: \"Traffic Sign Recognition with Multi-Scale Convolutional Networks\" by `Pierre Sermanet` and `Yann LeCun`\n\nHere in this architecture, the 1st stage's ouput is feed-forward to the classifier (could be considered as a 3rd stage).", "_____no_output_____" ] ], [ [ "#LeNet model\ninputs = keras.Input(shape=(32,32,3), name='image_in')\n\n#0 stage :conversion from normalized RGB [0..1] to HSV\nlayer_HSV = tf.image.rgb_to_hsv(inputs)\n\n#1st stage ___________________________________________________________\n#Convolution with ReLU activation\nlayer1_conv = keras.layers.Conv2D(256, kernel_size=(5,5), strides=1, activation='relu', padding='valid')(layer_HSV)\n#Average Pooling\nlayer1_maxpool = keras.layers.MaxPooling2D(pool_size=(2,2), strides=2, padding='valid')(layer1_conv)\n#Conv 1x1\nlayer1_conv1x1 = keras.layers.Conv2D(256, kernel_size=(14,14), strides=1, activation='relu', padding='valid')(layer1_maxpool)\n\n#2nd stage ___________________________________________________________\n#Convolution with ReLU activation\nlayer2_conv = keras.layers.Conv2D(64, kernel_size=(5,5), strides=1, activation='relu', padding='valid')(layer1_maxpool)\n#MaxPooling 2D\nlayer2_maxpool = keras.layers.MaxPooling2D(pool_size=(2,2), strides=2, padding='valid')(layer2_conv)\n#Conv 1x1\nlayer2_conv1x1 = keras.layers.Conv2D(512, kernel_size=(5,5), strides=1, activation='relu', padding='valid')(layer2_maxpool)\n\n#3rd stage | Classifier ______________________________________________\n#Concate\nlayer3_flatten_1 = keras.layers.Flatten()(layer1_conv1x1)\nlayer3_flatten_2 = keras.layers.Flatten()(layer2_conv1x1)\nlayer3_concat = keras.layers.Concatenate()([layer3_flatten_1, layer3_flatten_2])\n\n#Dense (fully-connected)\nlayer3_dense_1 = keras.layers.Dense(units=129, activation='relu', kernel_initializer=\"he_normal\")(layer3_concat)\n# layer3_dense_2 = keras.layers.Dense(units=129, activation='relu', kernel_initializer=\"he_normal\")(layer3_dense_1)\n#Dense (fully-connected) | logits for 43 categories (n_classes)\noutputs = keras.layers.Dense(units=n_classes)(layer3_dense_1)\n\nLeNet_Model = keras.Model(inputs, outputs, name=\"LeNet_Model_improved\")", "_____no_output_____" ], [ "#Plot model architecture\nLeNet_Model.summary()\nkeras.utils.plot_model(LeNet_Model, \"figures/LeNet_improved_HLS.png\", show_shapes=True)", "Model: \"LeNet_Model_improved\"\n__________________________________________________________________________________________________\nLayer (type) Output Shape Param # Connected to \n==================================================================================================\nimage_in (InputLayer) [(None, 32, 32, 3)] 0 \n__________________________________________________________________________________________________\ntf_op_layer_RGBToHSV (TensorFlo [(None, 32, 32, 3)] 0 image_in[0][0] \n__________________________________________________________________________________________________\nconv2d (Conv2D) (None, 28, 28, 256) 19456 tf_op_layer_RGBToHSV[0][0] \n__________________________________________________________________________________________________\nmax_pooling2d (MaxPooling2D) (None, 14, 14, 256) 0 conv2d[0][0] \n__________________________________________________________________________________________________\nconv2d_2 (Conv2D) (None, 10, 10, 64) 409664 max_pooling2d[0][0] \n__________________________________________________________________________________________________\nmax_pooling2d_1 (MaxPooling2D) (None, 5, 5, 64) 0 conv2d_2[0][0] \n__________________________________________________________________________________________________\nconv2d_1 (Conv2D) (None, 1, 1, 256) 12845312 max_pooling2d[0][0] \n__________________________________________________________________________________________________\nconv2d_3 (Conv2D) (None, 1, 1, 512) 819712 max_pooling2d_1[0][0] \n__________________________________________________________________________________________________\nflatten (Flatten) (None, 256) 0 conv2d_1[0][0] \n__________________________________________________________________________________________________\nflatten_1 (Flatten) (None, 512) 0 conv2d_3[0][0] \n__________________________________________________________________________________________________\nconcatenate (Concatenate) (None, 768) 0 flatten[0][0] \n flatten_1[0][0] \n__________________________________________________________________________________________________\ndense (Dense) (None, 129) 99201 concatenate[0][0] \n__________________________________________________________________________________________________\ndense_1 (Dense) (None, 50) 6500 dense[0][0] \n==================================================================================================\nTotal params: 14,199,845\nTrainable params: 14,199,845\nNon-trainable params: 0\n__________________________________________________________________________________________________\n" ] ], [ [ "### Input preprocessing\n\n#### Color-Space\nPierre Sermanet and Yann LeCun used YUV color space with almost of processings on Y-channel (Y stands for brightness, U and V stand for Chrominance). \n\n#### Normalization\nEach channel of an image is in uint8 scale (0-255), we will normalize each channel to 0-1. \nGenerally, we normalize data to get them center around -1 and 1, to prevent numrical errors due to many steps of matrix operation. Imagine that we have 255x255x255x255xk operation, it could give a huge numerical error if we just have a small error in k. ", "_____no_output_____" ] ], [ [ "import cv2\ndef input_normalization(X_in):\n X = numpy.float32(X_in/255.0) \n return X", "_____no_output_____" ], [ "# normalization of dataset\n# enhanced training dataset is added\nX_train_norm = input_normalization(X_train)\nX_valid_norm = input_normalization(X_valid)\nX_enhancedtrain_norm = input_normalization(images_enhancedsign)\n\n# one-hot matrix\ny_train_onehot = keras.utils.to_categorical(y_train, n_classes)\ny_valid_onehot = keras.utils.to_categorical(y_valid, n_classes)\ny_enhanced_onehot = keras.utils.to_categorical(y_enhancedsign, n_classes)", "_____no_output_____" ], [ "print(X_train_norm.shape)\nprint('{0:.4g}'.format(numpy.max(X_train_norm)))\nprint('{0:.3g}'.format(numpy.min(X_train_norm)))", "(34799, 32, 32, 3)\n1\n0\n" ], [ "print(X_enhancedtrain_norm.shape)\nprint('{0:.4g}'.format(numpy.max(X_enhancedtrain_norm)))\nprint('{0:.3g}'.format(numpy.min(X_enhancedtrain_norm)))", "(50, 32, 32, 3)\n1\n0\n" ] ], [ [ "### Training Pipeline\n_Optimizer : we use Adam optimizer, better than SDG (Stochastic Gradient Descent) \n_Loss function : Cross Entropy by category \n_Metrics : accuracy \n*learning rate 0.001 work well with our network, it's better to try with small laerning rate in the begining.", "_____no_output_____" ] ], [ [ "rate = 0.001\n\nLeNet_Model.compile(\n optimizer=keras.optimizers.Nadam(learning_rate = rate, beta_1=0.9, beta_2=0.999, epsilon=1e-07),\n loss=keras.losses.CategoricalCrossentropy(from_logits=True),\n metrics=[\"accuracy\"])", "_____no_output_____" ] ], [ [ "### Real-time data augmentation", "_____no_output_____" ] ], [ [ "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n\ndatagen_enhanced = ImageDataGenerator(\n rotation_range=30.0,\n zoom_range=0.5,\n width_shift_range=0.5,\n height_shift_range=0.5,\n featurewise_center=True,\n featurewise_std_normalization=True,\n horizontal_flip=False)\n\ndatagen_enhanced.fit(X_enhancedtrain_norm)", "_____no_output_____" ], [ "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n\ndatagen = ImageDataGenerator(\n rotation_range=15.0,\n zoom_range=0.2,\n width_shift_range=0.1,\n height_shift_range=0.1,\n featurewise_center=False,\n featurewise_std_normalization=False,\n horizontal_flip=False)\n\ndatagen.fit(X_train_norm)", "_____no_output_____" ] ], [ [ "### Train the Model", "_____no_output_____" ], [ "###### on standard training dataset", "_____no_output_____" ] ], [ [ "EPOCHS = 30\nBATCH_SIZE = 32\nSTEPS_PER_EPOCH = int(len(X_train_norm)/BATCH_SIZE)\n\nhistory_standard_HLS = LeNet_Model.fit(\n datagen.flow(X_train_norm, y_train_onehot, batch_size=BATCH_SIZE,shuffle=True),\n validation_data=(X_valid_norm, y_valid_onehot),\n shuffle=True,\n steps_per_epoch=STEPS_PER_EPOCH,\n epochs=EPOCHS)", "WARNING:tensorflow:sample_weight modes were coerced from\n ...\n to \n ['...']\nTrain for 1087 steps, validate on 4410 samples\nEpoch 1/30\n1087/1087 [==============================] - 310s 285ms/step - loss: 1.5436 - accuracy: 0.5221 - val_loss: 1.1918 - val_accuracy: 0.6120\nEpoch 2/30\n1087/1087 [==============================] - 305s 281ms/step - loss: 0.5978 - accuracy: 0.7973 - val_loss: 0.8420 - val_accuracy: 0.7415\nEpoch 3/30\n1087/1087 [==============================] - 304s 279ms/step - loss: 0.3351 - accuracy: 0.8877 - val_loss: 0.8672 - val_accuracy: 0.7873\nEpoch 4/30\n1087/1087 [==============================] - 304s 279ms/step - loss: 0.2341 - accuracy: 0.9236 - val_loss: 0.7044 - val_accuracy: 0.8091\nEpoch 5/30\n1087/1087 [==============================] - 304s 279ms/step - loss: 0.1861 - accuracy: 0.9395 - val_loss: 0.6862 - val_accuracy: 0.8245\nEpoch 6/30\n1087/1087 [==============================] - 304s 279ms/step - loss: 0.1643 - accuracy: 0.9459 - val_loss: 0.6359 - val_accuracy: 0.8508\nEpoch 7/30\n1087/1087 [==============================] - 308s 284ms/step - loss: 0.1401 - accuracy: 0.9541 - val_loss: 0.7359 - val_accuracy: 0.8365\nEpoch 8/30\n1087/1087 [==============================] - 316s 290ms/step - loss: 0.1286 - accuracy: 0.9591 - val_loss: 0.7600 - val_accuracy: 0.8488\nEpoch 9/30\n1087/1087 [==============================] - 307s 282ms/step - loss: 0.1214 - accuracy: 0.9617 - val_loss: 0.8064 - val_accuracy: 0.8442\nEpoch 10/30\n1087/1087 [==============================] - 306s 282ms/step - loss: 0.1143 - accuracy: 0.9632 - val_loss: 0.7421 - val_accuracy: 0.8528\nEpoch 11/30\n1087/1087 [==============================] - 306s 281ms/step - loss: 0.1024 - accuracy: 0.9677 - val_loss: 0.7401 - val_accuracy: 0.8676\nEpoch 12/30\n1087/1087 [==============================] - 305s 281ms/step - loss: 0.0954 - accuracy: 0.9702 - val_loss: 0.6620 - val_accuracy: 0.8762\nEpoch 13/30\n1087/1087 [==============================] - 306s 281ms/step - loss: 0.1052 - accuracy: 0.9678 - val_loss: 0.7756 - val_accuracy: 0.8617\nEpoch 14/30\n1087/1087 [==============================] - 307s 282ms/step - loss: 0.0925 - accuracy: 0.9714 - val_loss: 0.6852 - val_accuracy: 0.8624\nEpoch 15/30\n1087/1087 [==============================] - 312s 287ms/step - loss: 0.0891 - accuracy: 0.9716 - val_loss: 0.9627 - val_accuracy: 0.8481\nEpoch 16/30\n1087/1087 [==============================] - 310s 285ms/step - loss: 0.0932 - accuracy: 0.9721 - val_loss: 0.7544 - val_accuracy: 0.8637\nEpoch 17/30\n1087/1087 [==============================] - 310s 285ms/step - loss: 0.0862 - accuracy: 0.9751 - val_loss: 0.6120 - val_accuracy: 0.8789\nEpoch 18/30\n1087/1087 [==============================] - 306s 281ms/step - loss: 0.0776 - accuracy: 0.9765 - val_loss: 0.8133 - val_accuracy: 0.8442\nEpoch 19/30\n1087/1087 [==============================] - 305s 280ms/step - loss: 0.0774 - accuracy: 0.9771 - val_loss: 0.6836 - val_accuracy: 0.8807\nEpoch 20/30\n1087/1087 [==============================] - 304s 280ms/step - loss: 0.0822 - accuracy: 0.9755 - val_loss: 0.8578 - val_accuracy: 0.8689\nEpoch 21/30\n1087/1087 [==============================] - 303s 279ms/step - loss: 0.0810 - accuracy: 0.9757 - val_loss: 0.7974 - val_accuracy: 0.8785\nEpoch 22/30\n1087/1087 [==============================] - 303s 279ms/step - loss: 0.0689 - accuracy: 0.9792 - val_loss: 0.8230 - val_accuracy: 0.8522\nEpoch 23/30\n1087/1087 [==============================] - 305s 280ms/step - loss: 0.0785 - accuracy: 0.9757 - val_loss: 0.6689 - val_accuracy: 0.8916\nEpoch 24/30\n1087/1087 [==============================] - 304s 280ms/step - loss: 0.0722 - accuracy: 0.9783 - val_loss: 0.9752 - val_accuracy: 0.8628\nEpoch 25/30\n1087/1087 [==============================] - 303s 279ms/step - loss: 0.0693 - accuracy: 0.9793 - val_loss: 1.0134 - val_accuracy: 0.8583\nEpoch 26/30\n1087/1087 [==============================] - 304s 279ms/step - loss: 0.0753 - accuracy: 0.9786 - val_loss: 0.6858 - val_accuracy: 0.8943\nEpoch 27/30\n1087/1087 [==============================] - 304s 280ms/step - loss: 0.0661 - accuracy: 0.9811 - val_loss: 1.0159 - val_accuracy: 0.8762\nEpoch 28/30\n1087/1087 [==============================] - 303s 279ms/step - loss: 0.0700 - accuracy: 0.9795 - val_loss: 1.0779 - val_accuracy: 0.8639\nEpoch 29/30\n1087/1087 [==============================] - 303s 279ms/step - loss: 0.0731 - accuracy: 0.9785 - val_loss: 0.8449 - val_accuracy: 0.8796\nEpoch 30/30\n1087/1087 [==============================] - 303s 279ms/step - loss: 0.0627 - accuracy: 0.9823 - val_loss: 1.0047 - val_accuracy: 0.8710\n" ] ], [ [ "###### on enhanced training dataset", "_____no_output_____" ] ], [ [ "EPOCHS = 30\nBATCH_SIZE = 1\nSTEPS_PER_EPOCH = int(len(X_enhancedtrain_norm)/BATCH_SIZE)\n\nhistory_enhanced_HLS = LeNet_Model.fit(\n datagen_enhanced.flow(X_enhancedtrain_norm, y_enhanced_onehot, batch_size=BATCH_SIZE,shuffle=True), \n shuffle=True, #validation_data=(X_valid_norm, y_valid_onehot),\n steps_per_epoch=STEPS_PER_EPOCH,\n epochs=EPOCHS)", "WARNING:tensorflow:sample_weight modes were coerced from\n ...\n to \n ['...']\nTrain for 50 steps\nEpoch 1/30\n50/50 [==============================] - 11s 229ms/step - loss: 351.2083 - accuracy: 0.0400\nEpoch 2/30\n50/50 [==============================] - 11s 219ms/step - loss: 142.5382 - accuracy: 0.0000e+00\nEpoch 3/30\n50/50 [==============================] - 11s 218ms/step - loss: 44.1551 - accuracy: 0.0400\nEpoch 4/30\n50/50 [==============================] - 11s 217ms/step - loss: 14.9848 - accuracy: 0.0200\nEpoch 5/30\n50/50 [==============================] - 11s 217ms/step - loss: 131.3266 - accuracy: 0.0000e+00\nEpoch 6/30\n50/50 [==============================] - 11s 217ms/step - loss: 540.1987 - accuracy: 0.0200\nEpoch 7/30\n50/50 [==============================] - 11s 217ms/step - loss: 5.0112 - accuracy: 0.0000e+00\nEpoch 8/30\n50/50 [==============================] - 11s 217ms/step - loss: 16.5539 - accuracy: 0.0200\nEpoch 9/30\n50/50 [==============================] - 11s 217ms/step - loss: 25.3070 - accuracy: 0.0200\nEpoch 10/30\n50/50 [==============================] - 11s 217ms/step - loss: 11.1368 - accuracy: 0.0200\nEpoch 11/30\n50/50 [==============================] - 11s 218ms/step - loss: 4.1143 - accuracy: 0.0000e+00\nEpoch 12/30\n50/50 [==============================] - 11s 218ms/step - loss: 3.9304 - accuracy: 0.0200\nEpoch 13/30\n50/50 [==============================] - 11s 217ms/step - loss: 3.9299 - accuracy: 0.0200\nEpoch 14/30\n50/50 [==============================] - 11s 218ms/step - loss: 3.9296 - accuracy: 0.0200\nEpoch 15/30\n50/50 [==============================] - 11s 217ms/step - loss: 3.9291 - accuracy: 0.0200\nEpoch 16/30\n50/50 [==============================] - 11s 217ms/step - loss: 8.8657 - accuracy: 0.0200\nEpoch 17/30\n50/50 [==============================] - 11s 224ms/step - loss: 3.9287 - accuracy: 0.0200\nEpoch 18/30\n50/50 [==============================] - 12s 231ms/step - loss: 3.9284 - accuracy: 0.0200\nEpoch 19/30\n50/50 [==============================] - 11s 223ms/step - loss: 3.9280 - accuracy: 0.0200\nEpoch 20/30\n50/50 [==============================] - 11s 223ms/step - loss: 3.9278 - accuracy: 0.0200\nEpoch 21/30\n50/50 [==============================] - 11s 221ms/step - loss: 3.9275 - accuracy: 0.0200\nEpoch 22/30\n50/50 [==============================] - 11s 218ms/step - loss: 3.9273 - accuracy: 0.0200\nEpoch 23/30\n50/50 [==============================] - 11s 220ms/step - loss: 3.9095 - accuracy: 0.0200\nEpoch 24/30\n50/50 [==============================] - 11s 221ms/step - loss: 3.9269 - accuracy: 0.0200\nEpoch 25/30\n50/50 [==============================] - 11s 220ms/step - loss: 10.9869 - accuracy: 0.0200\nEpoch 26/30\n50/50 [==============================] - 11s 220ms/step - loss: 3.9265 - accuracy: 0.0200\nEpoch 27/30\n50/50 [==============================] - 11s 219ms/step - loss: 6.8765 - accuracy: 0.0200\nEpoch 28/30\n50/50 [==============================] - 11s 221ms/step - loss: 3.9259 - accuracy: 0.0200\nEpoch 29/30\n50/50 [==============================] - 11s 219ms/step - loss: 3.9259 - accuracy: 0.0200\nEpoch 30/30\n50/50 [==============================] - 11s 218ms/step - loss: 30.3131 - accuracy: 0.0200\n" ], [ "LeNet_Model.save(\"LeNet_enhanced_trainingdataset_HLS.h5\")", "_____no_output_____" ] ], [ [ "### Evaluate the Model\nWe will use the test dataset to evaluate classification accuracy.", "_____no_output_____" ] ], [ [ "#Normalize test dataset\nX_test_norm = input_normalization(X_test)\n#One-hot matrix\ny_test_onehot = keras.utils.to_categorical(y_test, n_classes)", "_____no_output_____" ], [ "#Load saved model\nreconstructed_LeNet_Model = keras.models.load_model(\"LeNet_enhanced_trainingdataset_HLS.h5\")\n\n#Evaluate and display the prediction\nresult = reconstructed_LeNet_Model.evaluate(X_test_norm,y_test_onehot)\ndict(zip(reconstructed_LeNet_Model.metrics_names, result))", "12630/12630 [==============================] - 26s 2ms/sample - loss: 3.9001 - accuracy: 0.0119s - loss: 3.9003 - accuracy: \n" ], [ "pickle.dump(history_enhanced_HLS.history, open( \"history_LeNet_enhanced_trainingdataset_enhanced_HLS.p\", \"wb\" ))\npickle.dump(history_standard_HLS.history, open( \"history_LeNet_enhanced_trainingdataset_standard_HLS.p\", \"wb\" ))", "_____no_output_____" ], [ "with open(\"history_LeNet_enhanced_trainingdataset_standard_HLS.p\", mode='rb') as f:\n history_ = pickle.load(f)", "_____no_output_____" ], [ "import matplotlib.pyplot as plt\n\n# Plot training error.\nprint('\\nPlot of training error over 30 epochs:')\nfig = plt.figure()\nplt.title('Training error')\nplt.ylabel('Cost')\nplt.xlabel('epoch')\nplt.plot(history_['loss'])\nplt.plot(history_['val_loss'])\n# plt.plot(history.history['loss'])\n# plt.plot(history.history['val_loss'])\nplt.legend(['train loss', 'val loss'], loc='upper right')\nplt.grid()\nplt.show()\nfig.savefig('figures/Training_loss_LeNet_enhanced_trainingdataset_standard_HLS.png', dpi=500)", "\nPlot of training error over 30 epochs:\n" ], [ "# Plot training error.\nprint('\\nPlot of training accuracy over 30 epochs:')\nfig = plt.figure()\nplt.title('Training accuracy')\nplt.ylabel('Accuracy')\nplt.ylim([0.4, 1])\nplt.xlabel('epoch')\nplt.plot(history_['accuracy'])\nplt.plot(history_['val_accuracy'])\nplt.legend(['training_accuracy', 'validation_accuracy'], loc='lower right')\nplt.grid()\nplt.show()\nfig.savefig('figures/Training_accuracy_LeNet_enhanced_trainingdataset_HLS.png', dpi=500)", "\nPlot of training accuracy over 30 epochs:\n" ] ], [ [ "### Prediction of test dataset with trained model\nWe will use the test dataset to test trained model's prediction of instances that it has never seen during training.", "_____no_output_____" ] ], [ [ "print(\"Test Set : {} samples\".format(len(X_test)))\nprint('n_classes : {}'.format(n_classes))\nX_test.shape", "Test Set : 12630 samples\nn_classes : 50\n" ], [ "#Normalize test dataset\nX_test_norm = input_normalization(X_test)\n#One-hot matrix\ny_test_onehot = keras.utils.to_categorical(y_test, n_classes)", "_____no_output_____" ], [ "#Load saved model\nreconstructed = keras.models.load_model(\"LeNet_enhanced_trainingdataset_HLS.h5\")\n\n#Evaluate and display the prediction\nprediction_performance = reconstructed.evaluate(X_test_norm,y_test_onehot)\ndict(zip(reconstructed.metrics_names, prediction_performance))", "12630/12630 [==============================] - 32s 3ms/sample - loss: 3.9001 - accuracy: 0.0119\n" ], [ "import matplotlib.pyplot as plot\n%matplotlib inline\nrows, cols = 4, 12\nfig, axes = plot.subplots(rows, cols)\nfor idx, ax in enumerate(axes.ravel()):\n if idx < n_classes_standard :\n X_test_of_class = X_test[y_test == idx]\n #X_train_0 = X_train_of_class[numpy.random.randint(len(X_train_of_class))]\n X_test_0 = X_test_of_class[0]\n ax.imshow(X_test_0)\n ax.axis('off')\n ax.set_title('{:02d}'.format(idx))\n plot.setp(ax.get_xticklabels(), visible=False)\n plot.setp(ax.get_yticklabels(), visible=False)\n else:\n ax.axis('off')\n#\nplot.draw()\nfig.savefig('figures/' + 'test_representative' + '.jpg', dpi=700)", "_____no_output_____" ], [ "#### Prediction for all instances inside the test dataset\ny_pred_proba = reconstructed.predict(X_test_norm)\ny_pred_class = y_pred_proba.argmax(axis=-1)\n\n### Showing prediction results for 10 first instances\nfor i, pred in enumerate(y_pred_class):\n if i <= 10: \n print('Image {} - Target = {}, Predicted = {}'.format(i, y_test[i], pred))\n else:\n break", "Image 0 - Target = 16, Predicted = 6\nImage 1 - Target = 1, Predicted = 6\nImage 2 - Target = 38, Predicted = 6\nImage 3 - Target = 33, Predicted = 6\nImage 4 - Target = 11, Predicted = 6\nImage 5 - Target = 38, Predicted = 6\nImage 6 - Target = 18, Predicted = 6\nImage 7 - Target = 12, Predicted = 6\nImage 8 - Target = 25, Predicted = 6\nImage 9 - Target = 35, Predicted = 6\nImage 10 - Target = 12, Predicted = 6\n" ] ], [ [ "We will display a confusion matrix on test dataset to figure out our error-rate. \n`X_test_norm` : test dataset \n`y_test` : test dataset ground truth labels \n`y_pred_class` : prediction labels on test dataset ", "_____no_output_____" ] ], [ [ "confusion_matrix = numpy.zeros([n_classes, n_classes])", "_____no_output_____" ] ], [ [ "#### confusion_matrix\n`column` : test dataset ground truth labels \n`row` : prediction labels on test dataset \n`diagonal` : incremented when prediction matches ground truth label ", "_____no_output_____" ] ], [ [ "for ij in range(len(X_test_norm)):\n if y_test[ij] == y_pred_class[ij]:\n confusion_matrix[y_test[ij],y_test[ij]] += 1\n else:\n confusion_matrix[y_pred_class[ij],y_test[ij]] -= 1", "_____no_output_____" ], [ "column_label = [' L % d' % x for x in range(n_classes)]\nrow_label = [' P % d' % x for x in range(n_classes)]", "_____no_output_____" ], [ "# Plot classe representatives\nimport matplotlib.pyplot as plot\n%matplotlib inline\nrows, cols = 1, 43\nfig, axes = plot.subplots(rows, cols)\nfor idx, ax in enumerate(axes.ravel()):\n if idx < n_classes :\n X_test_of_class = X_test[y_test == idx]\n X_test_0 = X_test_of_class[0]\n ax.imshow(X_test_0)\n plot.setp(ax.get_xticklabels(), visible=False)\n plot.setp(ax.get_yticklabels(), visible=False)\n # plot.tick_params(axis='both', which='both', bottom='off', top='off',\n # labelbottom='off', right='off', left='off', labelleft='off')\n ax.axis('off')\nplot.draw()\nfig.savefig('figures/' + 'label_groundtruth' + '.jpg', dpi=3500)", "_____no_output_____" ], [ "numpy.savetxt(\"confusion_matrix_LeNet_enhanced_trainingdataset_HLS.csv\", confusion_matrix, delimiter=\";\")", "_____no_output_____" ] ], [ [ "#### Thank to confusion matrix, we could identify where to enhance \n-[x] training dataset \n-[x] real-time data augmentation \n-[x] preprocessing \n\n*Extract of confusion matrix of classification on test dataset &#8595;* \n<img src=\"figures/confusion_matrix_LeNet_enhanced_trainingdataset_HLS.png\" alt=\"Drawing\" style=\"width: 750px;\"/>", "_____no_output_____" ], [ "### Prediction of new instances with trained model\nWe will use the test dataset to test trained model's prediction of instances that it has never seen during training.\nI didn't 'softmax' activation in the last layer of LeNet architecture, so the output prediction is logits. To have prediction confidence level, we can apply softmax function to output logits.", "_____no_output_____" ] ], [ [ "# load french traffic signs\nimport os\nimport cv2\nimport matplotlib.pyplot as plot\nimport numpy\ndir_frenchsign = 'french_traffic-signs-data'\nimages_frenchsign = [os.path.join(dir_frenchsign, f) for f in os.listdir(dir_frenchsign)]\nimages_frenchsign = numpy.array([cv2.cvtColor(cv2.imread(f), cv2.COLOR_BGR2RGB) for f in images_frenchsign])\n\n# plot new test images\nfig, axes = plot.subplots(3, int(len(images_frenchsign)/3))\nplot.title('French traffic signs')\nfor i, ax in enumerate(axes.ravel()):\n ax.imshow(images_frenchsign[i])\n ax.set_title('{}'.format(i))\n plot.setp(ax.get_xticklabels(), visible=False)\n plot.setp(ax.get_yticklabels(), visible=False)\n ax.set_xticks([]), ax.set_yticks([])\n ax.axis('off')\nplot.draw()\nfig.savefig('figures/' + 'french_sign' + '.jpg', dpi=700)", "_____no_output_____" ] ], [ [ "*Enhanced German traffic signs dataset &#8595;* \n<img src=\"figures/enhanced_training_dataset.png\" alt=\"Drawing\" style=\"width: 700px;\"/>", "_____no_output_____" ] ], [ [ "# manually label for these new images\ny_frenchsign = [13, 31, 29, 24, 26, 27, 33, 17, 15, 34, 12, 2, 2, 4, 2]\nn_classes = n_classes_enhanced\n\n# when a sign doesn't present in our training dataset, we'll try to find a enough 'similar' sign to label it.\n\n# image 2 : class 29 differed \n# image 3 : class 24, double-sens not existed \n# image 5 : class 27 differed\n# image 6 : class 33 not existed \n# image 7 : class 17, halte-péage not existed \n# image 8 : class 15, 3.5t limit not existed \n# image 9 : class 15, turn-left inhibition not existed\n# image 12 : class 2, ending of 50kmh speed-limit not existed\n# image 14 : class 2, 90kmh speed-limit not existed", "_____no_output_____" ] ], [ [ "#### it's really intersting that somes common french traffic signs are not present in INI German traffic signs dataset or differed\nWhatever our input - evenif it's not present in the training dataset, by using softmax activation our classififer can not say that 'this is a new traffic sign that it doesn't recognize' (sum of probability across all classes is 1), it's just try to find class that probably suit for the input.", "_____no_output_____" ] ], [ [ "#Normalize the dataset\nX_frenchsign_norm = input_normalization(images_frenchsign)\n#One-hot matrix\ny_frenchsign_onehot = keras.utils.to_categorical(y_frenchsign, n_classes)\n\n#Load saved model\nreconstructed = keras.models.load_model(\"LeNet_enhanced_trainingdataset_HLS.h5\")\n\n#Evaluate and display the prediction performance\nprediction_performance = reconstructed.evaluate(X_frenchsign_norm, y_frenchsign_onehot)\ndict(zip(reconstructed.metrics_names, prediction_performance))", "\r15/15 [==============================] - 0s 24ms/sample - loss: 3.9201 - accuracy: 0.0000e+00\n" ], [ "#### Prediction for all instances inside the test dataset\ny_pred_logits = reconstructed.predict(X_frenchsign_norm)\ny_pred_proba = tf.nn.softmax(y_pred_logits).numpy()\ny_pred_class = y_pred_proba.argmax(axis=-1)\n\n### Showing prediction results\nfor i, pred in enumerate(y_pred_class): \n print('Image {} - Target = {}, Predicted = {}'.format(i, y_frenchsign[i], pred))", "Image 0 - Target = 13, Predicted = 6\nImage 1 - Target = 31, Predicted = 6\nImage 2 - Target = 29, Predicted = 6\nImage 3 - Target = 24, Predicted = 6\nImage 4 - Target = 26, Predicted = 6\nImage 5 - Target = 27, Predicted = 6\nImage 6 - Target = 33, Predicted = 6\nImage 7 - Target = 17, Predicted = 6\nImage 8 - Target = 15, Predicted = 6\nImage 9 - Target = 34, Predicted = 6\nImage 10 - Target = 12, Predicted = 6\nImage 11 - Target = 2, Predicted = 6\nImage 12 - Target = 2, Predicted = 6\nImage 13 - Target = 4, Predicted = 6\nImage 14 - Target = 2, Predicted = 6\n" ] ], [ [ "*French traffic signs to classsify &#8595;* \n<img src=\"figures/french_sign_compare_german_INI_enhanced.jpg\" alt=\"Drawing\" style=\"width: 750px;\"/>", "_____no_output_____" ] ], [ [ "#### plot softmax probs along with traffic sign examples\nn_img = X_frenchsign_norm.shape[0]\nfig, axarray = plot.subplots(n_img, 2)\nplot.suptitle('Visualization of softmax probabilities', fontweight='bold')\nfor r in range(0, n_img):\n axarray[r, 0].imshow(numpy.squeeze(images_frenchsign[r]))\n axarray[r, 0].set_xticks([]), axarray[r, 0].set_yticks([])\n plot.setp(axarray[r, 0].get_xticklabels(), visible=False)\n plot.setp(axarray[r, 0].get_yticklabels(), visible=False)\n axarray[r, 1].bar(numpy.arange(n_classes), y_pred_proba[r])\n axarray[r, 1].set_ylim([0, 1])\n plot.setp(axarray[r, 1].get_yticklabels(), visible=False)\nplot.draw()\nfig.savefig('figures/' + 'french_sign_softmax_visuali_LeNet_enhanced_trainingdataset_HLS' + '.jpg', dpi=700)\n \nK = 3\n#### print top K predictions of the model for each example, along with confidence (softmax score) \nfor i in range(len(images_frenchsign)):\n print('Top {} model predictions for image {} (Target is {:02d})'.format(K, i, y_frenchsign[i]))\n top_3_idx = numpy.argsort(y_pred_proba[i])[-3:]\n top_3_values = y_pred_proba[i][top_3_idx] \n top_3_logits = y_pred_logits[i][top_3_idx] \n for k in range(K):\n print(' Prediction = {:02d} with probability {:.4f} (logit is {:.4f})'.format(top_3_idx[k], top_3_values[k], top_3_logits[k]))", "Top 3 model predictions for image 0 (Target is 13)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 1 (Target is 31)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 2 (Target is 29)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 3 (Target is 24)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 4 (Target is 26)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 5 (Target is 27)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 6 (Target is 33)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 7 (Target is 17)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 8 (Target is 15)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 9 (Target is 34)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 10 (Target is 12)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 11 (Target is 02)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 12 (Target is 02)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 13 (Target is 04)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\nTop 3 model predictions for image 14 (Target is 02)\n Prediction = 02 with probability 0.0250 (logit is 0.1806)\n Prediction = 31 with probability 0.0262 (logit is 0.2275)\n Prediction = 06 with probability 0.0292 (logit is 0.3377)\n" ] ], [ [ "*Visualization of softmax probabilities &#8595;* \n<img src=\"figures/french_sign_softmax_visuali_LeNet_enhanced_trainingdataset_HLS.jpg\" alt=\"Drawing\" style=\"width: 750px;\"/>", "_____no_output_____" ], [ "## Visualization of layers", "_____no_output_____" ] ], [ [ "### Import tensorflow and keras\nimport tensorflow as tf\nfrom tensorflow import keras\nfrom tensorflow.keras import Model\nimport matplotlib.pyplot as plot\n\nprint (\"TensorFlow version: \" + tf.__version__)", "TensorFlow version: 2.1.0\n" ], [ "# Load pickled data\nimport pickle\nimport numpy\n\ntraining_file = 'traffic-signs-data/train.p'\nwith open(training_file, mode='rb') as f:\n train = pickle.load(f)\n \nX_train, y_train = train['features'], train['labels'] # training dataset\nn_classes = len(numpy.unique(y_train))", "_____no_output_____" ], [ "import cv2\ndef input_normalization(X_in):\n X = numpy.float32(X_in/255.0) \n return X\n\n# normalization of dataset\nX_train_norm = input_normalization(X_train)\n\n# one-hot matrix\ny_train_onehot = keras.utils.to_categorical(y_train, n_classes)", "_____no_output_____" ], [ "#Load saved model\nreconstructed = keras.models.load_model(\"LeNet_enhanced_trainingdataset_HLS.h5\")\n\n#Build model for layer display\nlayers_output = [layer.output for layer in reconstructed.layers]\noutputs_model = Model(inputs=reconstructed.input, outputs=layers_output)\noutputs_history = outputs_model.predict(X_train_norm[900].reshape(1,32,32,3))", "_____no_output_____" ] ], [ [ "#### Display analized input", "_____no_output_____" ] ], [ [ "plot.imshow(X_train[900])", "_____no_output_____" ], [ "def display_layer(outputs_history, col_size, row_size, layer_index): \n activation = outputs_history[layer_index]\n activation_index = 0\n fig, ax = plot.subplots(row_size, col_size, figsize=(row_size*2.5,col_size*1.5))\n for row in range(0,row_size):\n for col in range(0,col_size):\n ax[row][col].axis('off')\n if activation_index < activation.shape[3]:\n ax[row][col].imshow(activation[0, :, :, activation_index]) # , cmap='gray' \n activation_index += 1", "_____no_output_____" ], [ "display_layer(outputs_history, 3, 2, 1)", "_____no_output_____" ], [ "display_layer(outputs_history, 8, 8, 2)", "_____no_output_____" ], [ "display_layer(outputs_history, 8, 8, 3)", "_____no_output_____" ], [ "display_layer(outputs_history, 8, 8, 4)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import escher\nimport escher.urls\nimport cobra\nimport cobra.test\nimport json\nimport os\nfrom IPython.display import HTML\nfrom copy import deepcopy", "_____no_output_____" ], [ "d = escher.urls.root_directory\nprint('Escher directory: %s' % d)", "Escher directory: /Users/joanrue/anaconda/lib/python2.7/site-packages\n" ] ], [ [ "### Embed an Escher map in an IPython notebook", "_____no_output_____" ] ], [ [ "escher.list_available_maps()", "_____no_output_____" ], [ "b = escher.Builder(map_name='e_coli_core.Core metabolism')\nb.display_in_notebook()", "_____no_output_____" ] ], [ [ "### Plot FBA solutions in Escher", "_____no_output_____" ] ], [ [ "model = cobra.io.load_json_model( \"iECW_1372.json\") # E coli metabolic model\nFBA_Solution = model.optimize() # FBA of the original model\n\nprint('Original Growth rate: %.9f' % FBA_Solution.f)\n\n", "Original Growth rate: 0.982478439\n" ], [ "b = escher.Builder(map_name='e_coli_core.Core metabolism',\n reaction_data=FBA_Solution.x_dict,\n # color and size according to the absolute value\n reaction_styles=['color', 'size', 'abs', 'text'],\n # change the default colors\n reaction_scale=[{'type': 'min', 'color': '#cccccc', 'size': 4},\n {'type': 'mean', 'color': '#0000dd', 'size': 20},\n {'type': 'max', 'color': '#ff0000', 'size': 40}],\n # only show the primary metabolites\n hide_secondary_metabolites=True,\n highlight_missing = True)\nb.display_in_notebook()\n#b.display_in_browser()", "_____no_output_____" ], [ "# MAP EDITION\nmodel_knockout = model.copy()\ncobra.manipulation.delete_model_genes(model_knockout, [\"ECW_m3223\"]) #ODC\ncobra.manipulation.delete_model_genes(model_knockout, [\"ECW_m0743\"]) #ODC\ncobra.manipulation.delete_model_genes(model_knockout, [\"ECW_m3196\"]) #Agmatinase. \nknockout_FBA_solution = model_knockout.optimize() # FBA of the knockout\n\nprint('Knockout Growth rate: %.9f' % knockout_FBA_solution.f)", "Knockout Growth rate: 0.982478439\n" ], [ "#PASS THE MODEL TO A NEW BUILDER\nb = escher.Builder(map_name='e_coli_core.Core metabolism',\n reaction_data=knockout_FBA_solution.x_dict,\n # color and size according to the absolute value\n reaction_styles=['color', 'size', 'abs', 'text'],\n # change the default colors\n reaction_scale=[{'type': 'min', 'color': '#cccccc', 'size': 4},\n {'type': 'mean', 'color': '#0000dd', 'size': 20},\n {'type': 'max', 'color': '#ff0000', 'size': 40}],\n # only show the primary metabolites\n hide_secondary_metabolites=True,\n highlight_missing = True)\nb.display_in_notebook()\n#b.display_in_browser()", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "x =2\nr = lambda x:x**2", "_____no_output_____" ], [ "r(x)", "_____no_output_____" ], [ "r1 = lambda x:x**2 if x>3 else None", "_____no_output_____" ], [ "r1(x)", "_____no_output_____" ], [ "r1(5)", "_____no_output_____" ], [ "#object\n\n#using class keyword\n#creating the attributes\n#creating the methods in class\n#learning about inheritence in python\n#learning about polymorphism", "_____no_output_____" ], [ "lt = [1,2,3,4]", "_____no_output_____" ], [ "lt.count(4)", "_____no_output_____" ], [ "print(type(1))\nprint(type([]))\nprint(type(()))\nprint(type({}))", "<class 'int'>\n<class 'list'>\n<class 'tuple'>\n<class 'dict'>\n" ], [ "#Create a new Object type ca;;ed test_sample\n\nclass TestSample:\n pass\n\n#Instance of the object \nx = TestSample()\nprint(type(x))", "<class '__main__.TestSample'>\n" ], [ "class Dog:\n def __init__(self,breed):\n self.breed = breed\n \ntuku = Dog(breed=\"Dalmi\") ", "_____no_output_____" ], [ "tuku.breed", "_____no_output_____" ], [ "#Init Method is constructor", "_____no_output_____" ], [ "#Self is keyword used for like this keyword ", "_____no_output_____" ] ], [ [ "# Methods", "_____no_output_____" ] ], [ [ "class ourCircle:\n pi = 3.14\n def __init__(self,radius=1):\n self.radius = radius\n self.area = self.getArea(radius)\n \n def setRadius(self,new_radius):\n self.radius = new_radius\n self = new_radius * new_radius * self.pi\n \n def getCircumference(self):\n return self.radius * self.pi * 2\n \n def getArea(self,radius):\n return radius * radius * self.pi\n", "_____no_output_____" ], [ "cir = ourCircle()", "_____no_output_____" ], [ "cir.getCircumference()", "_____no_output_____" ], [ "cir.getArea(2)", "_____no_output_____" ] ], [ [ "# Inheritance", "_____no_output_____" ] ], [ [ "class Animal:\n def __init__(self):\n print(\"Animal Object Cost Created\")\n def whoAmI(self):\n print(\"I am Animal Class\")\n def eat(self):\n print(\"I am eating\")", "_____no_output_____" ], [ "a = Animal()", "Animal Object Cost Created\n" ], [ "a.eat()", "I am eating\n" ], [ "class Man(Animal):\n def __init__(self):\n Animal.__init__(self) ", "_____no_output_____" ], [ "m = Man()", "Animal Object Cost Created\n" ], [ "m.eat()", "I am eating\n" ], [ "#Polymorphism", "_____no_output_____" ], [ "#Exception Handling", "_____no_output_____" ], [ "try:\n f =open('test','w')\n # f.write(\"rams\")\n f.read()\nexcept IOError:\n print('geting error')\nelse:\n print(\"Print CONETENT SUCCESS\")\n f.close()\n \nprint(\"Error Handle by try and catch\")", "geting error\nError Handle by try and catch\n" ], [ "try:\n a='ram'\n b=10\n print(a+b)\nexcept Exception as e:\n print(e.args)\n ", "('can only concatenate str (not \"int\") to str',)\n" ], [ "try:\n a='ram'\n b=10\n print(a+b)\nexcept Exception as e:\n print(e.args)\nfinally:\n print(\"i'm getting error\")", "('can only concatenate str (not \"int\") to str',)\ni'm getting error\n" ], [ "while False:\n print(\"rams\")\nelse:\n print(\"vijay\")", "vijay\n" ], [ "#pip install pylint", "_____no_output_____" ], [ "%%writefile simple.py\n\na=10\nprint(a)", "Writing simple.py\n" ], [ "! pylint simple.py", "************* Module simple\nsimple.py:2:1: C0326: Exactly one space required around assignment\na=10\n\n ^ (bad-whitespace)\nsimple.py:1:0: C0111: Missing module docstring (missing-docstring)\nsimple.py:2:0: C0103: Constant name \"a\" doesn't conform to UPPER_CASE naming style (invalid-name)\n\n------------------------------------\n\nYour code has been rated at -5.00/10\n\n\n\n" ], [ "%%writefile simple.py\n'''\nA very simple script\n'''\ndef fun():\n first = 1\n second = 2 \n print(first)\n print(second)\nfun() ", "Overwriting simple.py\n" ], [ "!pylint simple.py", "************* Module simple\nsimple.py:6:14: C0303: Trailing whitespace (trailing-whitespace)\nsimple.py:9:5: C0303: Trailing whitespace (trailing-whitespace)\nsimple.py:4:0: C0111: Missing function docstring (missing-docstring)\n\n------------------------------------------------------------------\n\nYour code has been rated at 5.00/10 (previous run: 3.33/10, +1.67)\n\n\n\n" ], [ "#UnitTest", "_____no_output_____" ], [ "%%writefile capitalize_text.py\ndef capitalize_test(text):\n return text.capitalize_t()", "Overwriting capitalize_text.py\n" ], [ "%%writefile test_file.py\nimport unittest\nimport capitalize_text\n\nclass TestCap(unittest.TestCase):\n \n def test_one_word(self):\n test = \"python\"\n result = capitalize_text.capitalize_test(text)\n self.assertEqual(result,'python')\n def test_multiple_words(self):\n text = \"my python\"\n result = capitalize_text.capitalize_test(text)\n self.assertEqual(result,'my python')\n \nif __name__ == '__main__':\n unittest.main()", "Overwriting test_file.py\n" ], [ "! python test_file.py", "EE\n======================================================================\nERROR: test_multiple_words (__main__.TestCap)\n----------------------------------------------------------------------\nTraceback (most recent call last):\n File \"test_file.py\", line 12, in test_multiple_words\n result = capitalize_text.capitalize_test(text)\n File \"C:\\Users\\Kumaran\\Documents\\FCS Python\\capitalize_text.py\", line 2, in capitalize_test\n return text.capitalize_t()\nAttributeError: 'str' object has no attribute 'capitalize_t'\n\n======================================================================\nERROR: test_one_word (__main__.TestCap)\n----------------------------------------------------------------------\nTraceback (most recent call last):\n File \"test_file.py\", line 8, in test_one_word\n result = capitalize_text.capitalize_test(text)\nNameError: name 'text' is not defined\n\n----------------------------------------------------------------------\nRan 2 tests in 0.001s\n\nFAILED (errors=2)\n" ], [ "! pylint capitalize_text.py", "************* Module capitalize_text\ncapitalize_text.py:1:0: C0111: Missing module docstring (missing-docstring)\ncapitalize_text.py:1:0: C0111: Missing function docstring (missing-docstring)\n\n-----------------------------------\n\nYour code has been rated at 0.00/10\n\n\n\n" ] ] ]

[ "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Simulating Grover's Search Algorithm with 2 Qubits", "_____no_output_____" ] ], [ [ "import numpy as np\nfrom matplotlib import pyplot as plt\n%matplotlib inline", "_____no_output_____" ] ], [ [ "Define the zero and one vectors\nDefine the initial state $\\psi$", "_____no_output_____" ] ], [ [ "zero = np.matrix([[1],[0]]);\none = np.matrix([[0],[1]]);\n\npsi = np.kron(zero,zero);\nprint(psi)", "[[1]\n [0]\n [0]\n [0]]\n" ] ], [ [ "Define the gates we will use:\n\n$\n\\text{Id} = \\begin{pmatrix} \n1 & 0 \\\\\n0 & 1 \n\\end{pmatrix},\n\\quad\nX = \\begin{pmatrix} \n0 & 1 \\\\\n1 & 0 \n\\end{pmatrix},\n\\quad\nZ = \\begin{pmatrix} \n1 & 0 \\\\\n0 & -1 \n\\end{pmatrix},\n\\quad\nH = \\frac{1}{\\sqrt{2}}\\begin{pmatrix} \n1 & 1 \\\\\n1 & -1 \n\\end{pmatrix},\n\\quad\n\\text{CNOT} = \\begin{pmatrix} \n1 & 0 & 0 & 0 \\\\\n0 & 1 & 0 & 0 \\\\ \n0 & 0 & 0 & 1 \\\\\n0 & 0 & 1 & 0\n\\end{pmatrix},\n\\quad\nCZ = (\\text{Id} \\otimes H) \\text{ CNOT } (\\text{Id} \\otimes H) \n$", "_____no_output_____" ] ], [ [ "Id = np.matrix([[1,0],[0,1]]);\nX = np.matrix([[0,1],[1,0]]);\nZ = np.matrix([[1,0],[0,-1]]);\nH = np.sqrt(0.5) * np.matrix([[1,1],[1,-1]]);\n\nCNOT = np.matrix([[1,0,0,0],[0,1,0,0],[0,0,0,1],[0,0,1,0]]);\nCZ = np.kron(Id,H).dot(CNOT).dot(np.kron(Id,H));\nprint(CZ)", "[[ 1. 0. 0. 0.]\n [ 0. 1. 0. 0.]\n [ 0. 0. 1. 0.]\n [ 0. 0. 0. -1.]]\n" ] ], [ [ "Define the oracle for Grover's algorithm (take search answer to be \"10\")\n\n$\n\\text{oracle} = \\begin{pmatrix} \n1 & 0 & 0 & 0 \\\\\n0 & 1 & 0 & 0 \\\\\n0 & 0 & -1 & 0 \\\\\n0 & 0 & 0 & 1\n\\end{pmatrix}\n= (Z \\otimes \\text{Id}) CZ\n$\n\nUse different combinations of $Z \\otimes \\text{Id}$ to change where search answer is. ", "_____no_output_____" ] ], [ [ "oracle = np.kron(Z,Id).dot(CZ);\nprint(oracle)", "[[ 1. 0. 0. 0.]\n [ 0. 1. 0. 0.]\n [ 0. 0. -1. 0.]\n [ 0. 0. 0. 1.]]\n" ] ], [ [ "Act the H gates on the input vector and apply the oracle ", "_____no_output_____" ] ], [ [ "psi0 = np.kron(H,H).dot(psi);\npsi1 = oracle.dot(psi0);\nprint(psi1)", "[[ 0.5]\n [ 0.5]\n [-0.5]\n [ 0.5]]\n" ] ], [ [ "Remember that when we measure the result (\"00\", \"01\", \"10\", \"11\") is chosen randomly with probabilities given by the vector elements squared.", "_____no_output_____" ] ], [ [ "print(np.multiply(psi1,psi1))", "[[ 0.25]\n [ 0.25]\n [ 0.25]\n [ 0.25]]\n" ] ], [ [ "There is no difference between any of the probabilities. It's still just a 25% chance of getting the right answer. \n\nWe need some of gates after the oracle before measuring to converge on the right answer. \n\nThese gates do the operation $W = \\frac{1}{2}\\begin{pmatrix} \n-1 & 1 & 1 & 1 \\\\\n1 & -1 & 1 & 1 \\\\\n1 & 1 & -1 & 1 \\\\\n1 & 1 & 1 & -1\n\\end{pmatrix}\n=\n(H \\otimes H)(Z \\otimes Z) CZ (H \\otimes H)\n$\n\nNotice that if the matrix W is multiplied by the vector after the oracle, W $\\frac{1}{2}\\begin{pmatrix} \n1 \\\\\n1 \\\\\n-1 \\\\\n1\n\\end{pmatrix} \n= \\begin{pmatrix} \n0 \\\\\n0 \\\\\n1 \\\\\n0\n\\end{pmatrix} $,\nevery vector element decreases, except the correct answer element which increases. This would be true if if we chose a different place for the search result originally.", "_____no_output_____" ] ], [ [ "W = np.kron(H,H).dot(np.kron(Z,Z)).dot(CZ).dot(np.kron(H,H));\nprint(W)", "[[-0.5 0.5 0.5 0.5]\n [ 0.5 -0.5 0.5 0.5]\n [ 0.5 0.5 -0.5 0.5]\n [ 0.5 0.5 0.5 -0.5]]\n" ], [ "psif = W.dot(psi1);\nprint(np.multiply(psif,psif))", "[[ 1.23259516e-32]\n [ 3.08148791e-33]\n [ 1.00000000e+00]\n [ 3.08148791e-33]]\n" ], [ "x = [0,1,2,3];\nxb = [0.25,1.25,2.25,3.25];\nlabels=['00', '01', '10', '11'];\nplt.axis([-0.5,3.5,-1.25,1.25]);\nplt.xticks(x,labels);\nplt.bar(x, np.ravel(psi0), 1/1.5, color=\"red\");\nplt.bar(xb, np.ravel(np.multiply(psi0,psi0)), 1/2., color=\"blue\");", "_____no_output_____" ], [ "labels=['00', '01', '10', '11'];\nplt.axis([-0.5,3.5,-1.25,1.25]);\nplt.xticks(x,labels);\nplt.bar(x, np.ravel(psi1), 1/1.5, color=\"red\");\nplt.bar(xb, np.ravel(np.multiply(psi1,psi1)), 1/2., color=\"blue\");", "_____no_output_____" ], [ "labels=['00', '01', '10', '11'];\nplt.axis([-0.5,3.5,-1.25,1.25]);\nplt.xticks(x,labels);\nplt.bar(x, np.ravel(psif), 1/1.5, color=\"red\");\nplt.bar(xb, np.ravel(np.multiply(psif,psif)), 1/2., color=\"blue\");", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<table class=\"ee-notebook-buttons\" align=\"left\">\n <td><a target=\"_blank\" href=\"https://github.com/giswqs/earthengine-py-notebooks/tree/master/Visualization/image_color_ramp.ipynb\"><img width=32px src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" /> View source on GitHub</a></td>\n <td><a target=\"_blank\" href=\"https://nbviewer.jupyter.org/github/giswqs/earthengine-py-notebooks/blob/master/Visualization/image_color_ramp.ipynb\"><img width=26px src=\"https://upload.wikimedia.org/wikipedia/commons/thumb/3/38/Jupyter_logo.svg/883px-Jupyter_logo.svg.png\" />Notebook Viewer</a></td>\n <td><a target=\"_blank\" href=\"https://mybinder.org/v2/gh/giswqs/earthengine-py-notebooks/master?filepath=Visualization/image_color_ramp.ipynb\"><img width=58px src=\"https://mybinder.org/static/images/logo_social.png\" />Run in binder</a></td>\n <td><a target=\"_blank\" href=\"https://colab.research.google.com/github/giswqs/earthengine-py-notebooks/blob/master/Visualization/image_color_ramp.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" /> Run in Google Colab</a></td>\n</table>", "_____no_output_____" ], [ "## Install Earth Engine API\nInstall the [Earth Engine Python API](https://developers.google.com/earth-engine/python_install) and [geehydro](https://github.com/giswqs/geehydro). The **geehydro** Python package builds on the [folium](https://github.com/python-visualization/folium) package and implements several methods for displaying Earth Engine data layers, such as `Map.addLayer()`, `Map.setCenter()`, `Map.centerObject()`, and `Map.setOptions()`.\nThe following script checks if the geehydro package has been installed. If not, it will install geehydro, which automatically install its dependencies, including earthengine-api and folium.", "_____no_output_____" ] ], [ [ "import subprocess\n\ntry:\n import geehydro\nexcept ImportError:\n print('geehydro package not installed. Installing ...')\n subprocess.check_call([\"python\", '-m', 'pip', 'install', 'geehydro'])", "_____no_output_____" ] ], [ [ "Import libraries", "_____no_output_____" ] ], [ [ "import ee\nimport folium\nimport geehydro", "_____no_output_____" ] ], [ [ "Authenticate and initialize Earth Engine API. You only need to authenticate the Earth Engine API once. ", "_____no_output_____" ] ], [ [ "try:\n ee.Initialize()\nexcept Exception as e:\n ee.Authenticate()\n ee.Initialize()", "_____no_output_____" ] ], [ [ "## Create an interactive map \nThis step creates an interactive map using [folium](https://github.com/python-visualization/folium). The default basemap is the OpenStreetMap. Additional basemaps can be added using the `Map.setOptions()` function. \nThe optional basemaps can be `ROADMAP`, `SATELLITE`, `HYBRID`, `TERRAIN`, or `ESRI`.", "_____no_output_____" ] ], [ [ "Map = folium.Map(location=[40, -100], zoom_start=4)\nMap.setOptions('HYBRID')", "_____no_output_____" ] ], [ [ "## Add Earth Engine Python script ", "_____no_output_____" ] ], [ [ "# Load SRTM Digital Elevation Model data.\nimage = ee.Image('CGIAR/SRTM90_V4');\n\n# Define an SLD style of discrete intervals to apply to the image.\nsld_intervals = \\\n '<RasterSymbolizer>' + \\\n '<ColorMap type=\"intervals\" extended=\"false\" >' + \\\n '<ColorMapEntry color=\"#0000ff\" quantity=\"0\" label=\"0\"/>' + \\\n '<ColorMapEntry color=\"#00ff00\" quantity=\"100\" label=\"1-100\" />' + \\\n '<ColorMapEntry color=\"#007f30\" quantity=\"200\" label=\"110-200\" />' + \\\n '<ColorMapEntry color=\"#30b855\" quantity=\"300\" label=\"210-300\" />' + \\\n '<ColorMapEntry color=\"#ff0000\" quantity=\"400\" label=\"310-400\" />' + \\\n '<ColorMapEntry color=\"#ffff00\" quantity=\"1000\" label=\"410-1000\" />' + \\\n '</ColorMap>' + \\\n '</RasterSymbolizer>';\n\n# Define an sld style color ramp to apply to the image.\nsld_ramp = \\\n '<RasterSymbolizer>' + \\\n '<ColorMap type=\"ramp\" extended=\"false\" >' + \\\n '<ColorMapEntry color=\"#0000ff\" quantity=\"0\" label=\"0\"/>' + \\\n '<ColorMapEntry color=\"#00ff00\" quantity=\"100\" label=\"100\" />' + \\\n '<ColorMapEntry color=\"#007f30\" quantity=\"200\" label=\"200\" />' + \\\n '<ColorMapEntry color=\"#30b855\" quantity=\"300\" label=\"300\" />' + \\\n '<ColorMapEntry color=\"#ff0000\" quantity=\"400\" label=\"400\" />' + \\\n '<ColorMapEntry color=\"#ffff00\" quantity=\"500\" label=\"500\" />' + \\\n '</ColorMap>' + \\\n '</RasterSymbolizer>';\n\n# Add the image to the map using both the color ramp and interval schemes.\nMap.setCenter(-76.8054, 42.0289, 8);\nMap.addLayer(image.sldStyle(sld_intervals), {}, 'SLD intervals');\nMap.addLayer(image.sldStyle(sld_ramp), {}, 'SLD ramp');", "_____no_output_____" ] ], [ [ "## Display Earth Engine data layers ", "_____no_output_____" ] ], [ [ "Map.setControlVisibility(layerControl=True, fullscreenControl=True, latLngPopup=True)\nMap", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "from sklearn.naive_bayes import MultinomialNB", "_____no_output_____" ], [ "modelo = MultinomialNB()", "_____no_output_____" ], [ "futebol1 = ['Castanho', 'Casado', 'Sexo Masculino', 'Cabelo Curto'];futebol2 = ['Azul', 'Casado', 'Sexo Masculino', 'Cabelo Curto']; futebol3 = ['Azul', 'Casado', 'Sexo Feminino', 'Cabelo Longo']; futebol4 = ['Azul', 'Não casado', 'Sexo Masculino', 'Cabelo Curto']; futebol5 = ['Azul','Não casado','Sexo Masculino','Cabelo Longo']", "_____no_output_____" ], [ "aerobica1 = ['Castanho', 'Casado', 'Sexo Feminino', 'Cabelo Curto'];aerobica2 = ['Castanho', 'Não casado', 'Sexo Feminino', 'Cabelo Curto']; aerobica3 = ['Castanho', 'Não casado', 'Sexo Feminino', 'Cabelo Longo']", "_____no_output_____" ], [ "misterioso1 = ['Castanho','Casado','Sexo Masculino','Cabelo Curto'];misterioso2 = ['Castanho','Não casado','Sexo Masculino','Cabelo Longo'];misterioso3 = ['Azul', 'Não casado', 'Sexo Feminino', 'Cabelo Longo'];misterioso4 = ['Azul', 'Casado', 'Sexo Masculino', 'Cabelo Longo']", "_____no_output_____" ], [ "dados = [futebol1, futebol2, futebol3, futebol4, futebol5, aerobica1, aerobica2, aerobica3]", "_____no_output_____" ], [ "marcacoes = [1,1,1,1,1,-1,-1,-1]", "_____no_output_____" ], [ "modelo.fit(dados, marcacoes)", "_____no_output_____" ], [ "print(modelo.predict([misterioso1, misterioso2, misterioso3, misterioso4]))", "[1 1 1 1]\n" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Ingeniería de Características", "_____no_output_____" ], [ "En las clases previas vimos las ideas fundamentales de machine learning, pero todos los ejemplos asumían que ya teníamos los datos numéricos en un formato ordenado de tamaño ``[n_samples, n_features]``.\nEn la realidad son raras las ocasiones en que los datos vienen así, _llegar y llevar_.\nCon esto en mente, uno de los pasos más importantes en la práctica de machine learging es la _ingeniería de características_ (_feature engineering_), que es tomar cualquier información que tengas sobre tu problema y convertirla en números con los que construirás tu matriz de características.\n\nEn esta sección veremos dos ejemplos comunes de _tareas_ de ingeniería de características: cómo representar _datos categóricos_ y cómo representar _texto_. \nOtras características más avanzandas, como el procesamiento de imágenes, quedarán para el fin del curso.\n\nAdicionalmente, discutiremos _características derivadas_ para incrementar la complejidad del modelo y la _imputación_ de datos perdidos.\nEn ocasiones este proceso se conoce como _vectorización_, ya que se refiere a convertir datos arbitrarios en vectores bien definidos.", "_____no_output_____" ], [ "## Características Categóricas\n\nUn tipo común de datos no numéricos son los datos _categóricos_.\nPor ejemplo, imagina que estás explorando datos de precios de propiedad, y junto a variables numéricas como precio (_price_) y número de habitaciones (_rooms_), también tienes información del barrio (_neighborhood_) de cada propiedad.\nPor ejemplo, los datos podrían verse así:", "_____no_output_____" ] ], [ [ "data = [\n {'price': 850000, 'rooms': 4, 'neighborhood': 'Queen Anne'},\n {'price': 700000, 'rooms': 3, 'neighborhood': 'Fremont'},\n {'price': 650000, 'rooms': 3, 'neighborhood': 'Wallingford'},\n {'price': 600000, 'rooms': 2, 'neighborhood': 'Fremont'}\n]", "_____no_output_____" ] ], [ [ "Podrías estar tentade a codificar estos datos directamente con un mapeo numérico:", "_____no_output_____" ] ], [ [ "{'Queen Anne': 1, 'Fremont': 2, 'Wallingford': 3};", "_____no_output_____" ] ], [ [ "Resulta que esto no es una buena idea. En Scikit-Learn, y en general, los modelos asumen que los datos numéricos reflejan cantidades algebraicas.\nUsar un mapeo así implica, por ejemplo, que *Queen Anne < Fremont < Wallingford*, o incluso que *Wallingford - Queen Anne = Fremont*, lo que no tiene mucho sentido.\n\nUna técnica que funciona en estas situaciones es _codificación caliente_ (_one-hot encoding_), que crea columnas numéricas que indican la presencia o ausencia de la categoría correspondiente, con un valor de 1 o 0 respectivamente.\nCuando tus datos son una lista de diccionarios, la clase ``DictVectorizer`` se encarga de la codificación por ti:", "_____no_output_____" ] ], [ [ "from sklearn.feature_extraction import DictVectorizer\nvec = DictVectorizer(sparse=False, dtype=int)\nvec.fit_transform(data)", "_____no_output_____" ] ], [ [ "Nota que la característica `neighborhood` se ha expandido en tres columnas separadas, representando las tres etiquetas de barrio, y que cada fila tiene un 1 en la columna asociada al barrio respectivo.\nTeniendo los datos codificados de esa manera, se puede proceder a ajustar un modelo en Scikit-Learn.\n\nPara ver el significado de cada columna se puede hacer lo siguiente:", "_____no_output_____" ] ], [ [ "vec.get_feature_names()", "_____no_output_____" ] ], [ [ "Hay una clara desventaja en este enfoque: si las categorías tienen muchos valores posibles, el dataset puede crecer demasiado.\nSin embargo, como los datos codificados contienen principalmente ceros, una matriz dispersa puede ser una solucion eficiente:", "_____no_output_____" ] ], [ [ "vec = DictVectorizer(sparse=True, dtype=int)\nvec.fit_transform(data)", "_____no_output_____" ] ], [ [ "Varios (pero no todos) de los estimadores en Scikit-Learn aceptan entradas dispersas. ``sklearn.preprocessing.OneHotEncoder`` y ``sklearn.feature_extraction.FeatureHasher`` son dos herramientas adicionales que permiten trabajar con este tipo de características.", "_____no_output_____" ], [ "## Texto\n\nOtra necesidad común es convertir texto en una serie de números que representen su contenido.\nPor ejemplo, mucho del análisis automático del contenido generado en redes sociales depende de alguna manera de codificar texto como números.\nUno de los métodos más simples es a través de _conteo de palabras_ (_word counts_): tomas cada pedazo del texto, cuentas las veces que aparece cada palabra en él, y pones los resultados en una table.\n\nPor ejemplo, considera las siguientes tres frases:", "_____no_output_____" ] ], [ [ "sample = ['problem of evil',\n 'evil queen',\n 'horizon problem']", "_____no_output_____" ] ], [ [ "Para vectorizar estos datos construiríamos una columna para las palabras \"problem,\" \"evil,\", \"horizon,\" etc.\nHacer esto a mano es posible, pero nos podemos ahorrar el tedio utilizando el ``CountVectorizer`` de Scikit-Learn:", "_____no_output_____" ] ], [ [ "from sklearn.feature_extraction.text import CountVectorizer\n\nvec = CountVectorizer()\nX = vec.fit_transform(sample)\nX", "_____no_output_____" ] ], [ [ "El resultado es una matriz dispersa que contiene cada vez que aparece cada palabra en los textos. Para inspeccionarlo fácilmente podemos convertir esto en un ``DataFrame``:", "_____no_output_____" ] ], [ [ "import pandas as pd\npd.DataFrame(X.toarray(), columns=vec.get_feature_names())", "_____no_output_____" ] ], [ [ "Todavía falta algo. Este enfoque puede tener problemas: el conteo de palabras puede hacer que algunas características pesen más que otras debido a la frecuencia con la que utilizamos las palabras, y esto puede ser sub-óptimo en algunos algoritmos de clasificación.\nUna manera de considerar esto es utilizar el modelo _frecuencia de términos-frecuencia inversa de documents_ (_TF-IDF_), que da peso a las palabras de acuerdo a qué tan frecuentemente aparecen en los documentos, pero también qué tan únicas son para cada documento.\nLa sintaxis para aplicar TF-IDF es similar a la que hemos visto antes:", "_____no_output_____" ] ], [ [ "from sklearn.feature_extraction.text import TfidfVectorizer\nvec = TfidfVectorizer()\nX = vec.fit_transform(sample)\npd.DataFrame(X.toarray(), columns=vec.get_feature_names())", "_____no_output_____" ] ], [ [ "Esto lo veremos en más detalle en la clase de Naive Bayes.", "_____no_output_____" ], [ "## Características Derivadas\n\nOtro tipo útil de característica es aquella derivada matemáticamente desde otras características en los datos de entrada.\nVimos un ejemplo en la clase de Hiperparámetros cuando construimos características polinomiales desde los datos.\nVimos que se puede convertir una regresión lineal en una polinomial sin usar un modelo distinto, sino que transformando los datos de entrada.\nEsto es conocido como _función de regresión base_ (_basis function regression_), y lo exploraremos en la clase de Regresión Lineal.\n\nPor ejemplo, es claro que los siguientes datos no se pueden describir por una línea recta:", "_____no_output_____" ] ], [ [ "%matplotlib inline\n%config InlineBackend.figure_format = 'retina'\n%matplotlib inline\nimport numpy as np\nimport matplotlib.pyplot as plt\n\nx = np.array([1, 2, 3, 4, 5])\ny = np.array([4, 2, 1, 3, 7])\nplt.scatter(x, y);", "_____no_output_____" ] ], [ [ "Si ajustamos una recta a los datos usando ``LinearRegression`` obtendremos un resultado óptimo:", "_____no_output_____" ] ], [ [ "from sklearn.linear_model import LinearRegression\nX = x[:, np.newaxis]\nmodel = LinearRegression().fit(X, y)\nyfit = model.predict(X)\nplt.scatter(x, y)\nplt.plot(x, yfit);", "_____no_output_____" ] ], [ [ "Es óptimo, pero también queda claro que necesitamos un modelo más sofisticado para describir la relació entre $x$ e $y$.\n\nUna manera de lograrlo es transformando los datos, agregando columnas o características adicionales que le den más flexibilidad al modelo. Por ejemplo, podemos agregar características polinomiales de la siguiente forma:", "_____no_output_____" ] ], [ [ "from sklearn.preprocessing import PolynomialFeatures\npoly = PolynomialFeatures(degree=3, include_bias=False)\nX2 = poly.fit_transform(X)\nprint(X2)", "[[ 1. 1. 1.]\n [ 2. 4. 8.]\n [ 3. 9. 27.]\n [ 4. 16. 64.]\n [ 5. 25. 125.]]\n" ] ], [ [ "Esta matriz de características _derivada_ tiene una columna que representa a $x$, una segunda columna que representa $x^2$, y una tercera que representa $x^3$.\nCalcular una regresión lineal en esta entrada da un ajuste más cercano a nuestros datos:", "_____no_output_____" ] ], [ [ "model = LinearRegression().fit(X2, y)\nyfit = model.predict(X2)\nplt.scatter(x, y)\nplt.plot(x, yfit);", "_____no_output_____" ] ], [ [ "La idea de mejorar un modelo sin cambiarlo, sino que transformando la entrada que recibe, es fundamental para muchas de las técnicas de machine learning más poderosas.\nExploraremos más esta idea en la clase de Regresión Lineal. \nEste camino es motivante y se puede generalizar con las técnicas conocidas como _métodos de kernel_, que exploraremos en la clase de _Support Vector Machines_ (SVM).", "_____no_output_____" ], [ "## Imputación de Datos Faltantes\n\nUna necesidad común en la ingeniería de características es la manipulación de datos faltantes.\nEn clases anteriores es posible que hayan visto el valor `NaN` en un `DataFrame`, utilizado para marcar valores que no existen.\nPor ejemplo, podríamos tener un dataset que se vea así:", "_____no_output_____" ] ], [ [ "from numpy import nan\nX = np.array([[ nan, 0, 3 ],\n [ 3, 7, 9 ],\n [ 3, 5, 2 ],\n [ 4, nan, 6 ],\n [ 8, 8, 1 ]])\ny = np.array([14, 16, -1, 8, -5])", "_____no_output_____" ] ], [ [ "Antes de aplicar un modelo a estos datos necesitamos reemplazar esos datos faltantes con algún valor apropiado de relleno.\nEsto es conocido como _imputación_ de valores faltantes, y las estrategias para hacerlo varían desde las más simples (como rellenar con el promedio de cada columna) hasta las más sofisticadas (como completar la matrix con un modelo robusto para esos casos). Estos últimos enfoques suelen ser específicos para cada aplicación, así que no los veremos en el curso.\n\nLa clase `Imputer` de Scikit-Learn provee en enfoque base de imputación que calcula el promedio, la media, o el valor más frecuente:", "_____no_output_____" ] ], [ [ "from sklearn.preprocessing import Imputer\nimp = Imputer(strategy='mean')\nX2 = imp.fit_transform(X)\nX2", "_____no_output_____" ] ], [ [ "Como vemos, al aplicar el imputador los dos valores que faltaban fueron reemplazados por el promedio de los valores presentes en las columnas respectivas. \nAhora que tenemos una matriz sin valores faltantes, podemos usarla con la instancia de un modelo, en este caso, una regresión lineal:", "_____no_output_____" ] ], [ [ "model = LinearRegression().fit(X2, y)\nmodel.predict(X2)", "_____no_output_____" ] ], [ [ "## Cadena de Procesamiento (_Pipeline_)\n\nConsiderando los ejemplos que hemos visto, es posible que sea tedioso hacer cada una de estas transformaciones a mano. En ocasiones querremos automatizar la cadena de procesamiento para un modelo. Imagina una secuencia como la siguiente:\n\n1. Imputar valores usando el promedio.\n2. Transformar las características incluyendo un factor cuadrático.\n3. Ajustar una regresión lineal.\n\nPara encadenar estas etapas Scikit-Learn provee una clase ``Pipeline``, que se usa como sigue:", "_____no_output_____" ] ], [ [ "from sklearn.pipeline import make_pipeline\n\nmodel = make_pipeline(Imputer(strategy='mean'),\n PolynomialFeatures(degree=2),\n LinearRegression())", "_____no_output_____" ] ], [ [ "Esta cadena o _pipeline_ se ve y actúa como un objeto estándar de Scikit-Learn, por lo que podemos utilizarla en todo lo que hemos visto hasta ahora que siga la receta de uso de Scikit-Learn.", "_____no_output_____" ] ], [ [ "model.fit(X, y) # X con valores faltantes\nprint(y)\nprint(model.predict(X))", "[14 16 -1 8 -5]\n[14. 16. -1. 8. -5.]\n" ] ], [ [ "Todos los pasos del modelo se aplican de manera automática.\n\n¡Ojo! Por simplicidad hemos aplicado el modelo a los mismos datos con los que lo hemos entrenado, por eso el resultado es perfecto (vean el material de la clase pasada para recordar por qué esto no es un buen criterio para evaluar el modelo).\n\nEn las próximas clases seguiremos utilizando _Pipelines_ para estructurar nuestro análisis.", "_____no_output_____" ], [ "\n\nEste notebook contiene un extracto del libro [Python Data Science Handbook](http://shop.oreilly.com/product/0636920034919.do) por Jake VanderPlas; el contenido también está disponible en [GitHub](https://github.com/jakevdp/PythonDataScienceHandbook).\n\nEl texto se distribuye bajo una licencia [CC-BY-NC-ND](https://creativecommons.org/licenses/by-nc-nd/3.0/us/legalcode), y el código se distribuye bajo la licencia [MIT](https://opensource.org/licenses/MIT). Si te parece que este contenido es útil, por favor considera apoyar el trabajo [comprando el libro](http://shop.oreilly.com/product/0636920034919.do).\n\nTraducción al castellano por [Eduardo Graells-Garrido](http://datagramas.cl), liberada bajo las mismas condiciones.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Pre-training VGG16 for Distillation", "_____no_output_____" ] ], [ [ "import torch \nimport torch.nn as nn\n\nfrom src.data.dataset import get_dataloader\n\nimport torchvision.transforms as transforms\n\nimport numpy as np\nimport matplotlib.pyplot as plt", "_____no_output_____" ], [ "DEVICE = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\nprint(DEVICE)", "cuda\n" ], [ "SEED = 0\n\nBATCH_SIZE = 32\nLR = 5e-4\nNUM_EPOCHES = 25", "_____no_output_____" ], [ "np.random.seed(SEED)\ntorch.manual_seed(SEED)\ntorch.cuda.manual_seed(SEED)\ntorch.backends.cudnn.benchmark = False\ntorch.backends.cudnn.deterministic = True", "_____no_output_____" ] ], [ [ "## Preprocessing", "_____no_output_____" ] ], [ [ "transform = transforms.Compose([\n transforms.RandomHorizontalFlip(),\n #transforms.RandomVerticalFlip(),\n transforms.ToTensor(),\n transforms.Normalize((0.5,), (0.5,))\n])", "_____no_output_____" ], [ "train_loader, val_loader, test_loader = get_dataloader(\"./data/CIFAR10/\", BATCH_SIZE)", "Files already downloaded and verified\nFiles already downloaded and verified\n" ] ], [ [ "## Model", "_____no_output_____" ] ], [ [ "from src.models.model import VGG16_classifier", "_____no_output_____" ], [ "classes = 10\nhidden_size = 512\ndropout = 0.3\n\nmodel = VGG16_classifier(classes, hidden_size, preprocess_flag=False, dropout=dropout).to(DEVICE)\nmodel", "_____no_output_____" ], [ "for img, label in train_loader:\n img = img.to(DEVICE)\n label = label.to(DEVICE)\n \n print(\"Input Image Dimensions: {}\".format(img.size()))\n print(\"Label Dimensions: {}\".format(label.size()))\n print(\"-\"*100)\n \n out = model(img)\n \n print(\"Output Dimensions: {}\".format(out.size()))\n break", "Input Image Dimensions: torch.Size([32, 3, 32, 32])\nLabel Dimensions: torch.Size([32])\n----------------------------------------------------------------------------------------------------\n" ] ], [ [ "## Training", "_____no_output_____" ] ], [ [ "criterion = nn.CrossEntropyLoss()\noptimizer = torch.optim.Adam(params=model.parameters(), lr=LR)", "_____no_output_____" ], [ "loss_hist = {\"train accuracy\": [], \"train loss\": [], \"val accuracy\": []}\n\nfor epoch in range(1, NUM_EPOCHES+1):\n model.train()\n \n epoch_train_loss = 0\n \n y_true_train = []\n y_pred_train = []\n \n for batch_idx, (img, labels) in enumerate(train_loader):\n img = img.to(DEVICE)\n labels = labels.to(DEVICE)\n \n preds = model(img)\n \n loss = criterion(preds, labels)\n optimizer.zero_grad()\n loss.backward()\n optimizer.step()\n \n y_pred_train.extend(preds.detach().argmax(dim=-1).tolist())\n y_true_train.extend(labels.detach().tolist())\n \n epoch_train_loss += loss.item()\n\n with torch.no_grad():\n model.eval()\n\n y_true_test = []\n y_pred_test = []\n\n for batch_idx, (img, labels) in enumerate(val_loader):\n img = img.to(DEVICE)\n label = label.to(DEVICE)\n\n preds = model(img)\n\n y_pred_test.extend(preds.detach().argmax(dim=-1).tolist())\n y_true_test.extend(labels.detach().tolist())\n\n test_total_correct = len([True for x, y in zip(y_pred_test, y_true_test) if x==y])\n test_total = len(y_pred_test)\n test_accuracy = test_total_correct * 100 / test_total\n \n loss_hist[\"train loss\"].append(epoch_train_loss)\n\n total_correct = len([True for x, y in zip(y_pred_train, y_true_train) if x==y])\n total = len(y_pred_train)\n accuracy = total_correct * 100 / total\n \n loss_hist[\"train accuracy\"].append(accuracy)\n loss_hist[\"val accuracy\"].append(test_accuracy)\n \n print(\"-------------------------------------------------\")\n print(\"Epoch: {} Train mean loss: {:.8f}\".format(epoch, epoch_train_loss))\n print(\" Train Accuracy%: \", accuracy, \"==\", total_correct, \"/\", total)\n print(\" Validation Accuracy%: \", test_accuracy, \"==\", test_total_correct, \"/\", test_total)\n print(\"-------------------------------------------------\")", "-------------------------------------------------\nEpoch: 1 Train mean loss: 2339.36016971\n Train Accuracy%: 39.394 == 19697 / 50000\n Validation Accuracy%: 58.94 == 2947 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 2 Train mean loss: 1438.08532357\n Train Accuracy%: 69.244 == 34622 / 50000\n Validation Accuracy%: 74.02 == 3701 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 3 Train mean loss: 1072.54530132\n Train Accuracy%: 78.84 == 39420 / 50000\n Validation Accuracy%: 72.7 == 3635 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 4 Train mean loss: 907.63304938\n Train Accuracy%: 82.384 == 41192 / 50000\n Validation Accuracy%: 79.96 == 3998 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 5 Train mean loss: 761.87718780\n Train Accuracy%: 85.84 == 42920 / 50000\n Validation Accuracy%: 78.86 == 3943 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 6 Train mean loss: 718.01751248\n Train Accuracy%: 86.67 == 43335 / 50000\n Validation Accuracy%: 80.56 == 4028 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 7 Train mean loss: 648.26146014\n Train Accuracy%: 87.85 == 43925 / 50000\n Validation Accuracy%: 80.4 == 4020 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 8 Train mean loss: 562.79290721\n Train Accuracy%: 89.456 == 44728 / 50000\n Validation Accuracy%: 82.54 == 4127 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 9 Train mean loss: 473.16751064\n Train Accuracy%: 91.106 == 45553 / 50000\n Validation Accuracy%: 80.64 == 4032 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 10 Train mean loss: 464.82368950\n Train Accuracy%: 91.514 == 45757 / 50000\n Validation Accuracy%: 83.52 == 4176 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 11 Train mean loss: 467.99637549\n Train Accuracy%: 91.622 == 45811 / 50000\n Validation Accuracy%: 83.34 == 4167 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 12 Train mean loss: 393.67967687\n Train Accuracy%: 93.106 == 46553 / 50000\n Validation Accuracy%: 81.48 == 4074 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 13 Train mean loss: 472.65135630\n Train Accuracy%: 91.76 == 45880 / 50000\n Validation Accuracy%: 82.14 == 4107 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 14 Train mean loss: 547.69220992\n Train Accuracy%: 90.29 == 45145 / 50000\n Validation Accuracy%: 83.0 == 4150 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 15 Train mean loss: 361.64964305\n Train Accuracy%: 93.574 == 46787 / 50000\n Validation Accuracy%: 83.78 == 4189 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 16 Train mean loss: 363.55912426\n Train Accuracy%: 93.738 == 46869 / 50000\n Validation Accuracy%: 82.42 == 4121 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 17 Train mean loss: 370.34851927\n Train Accuracy%: 93.69 == 46845 / 50000\n Validation Accuracy%: 84.9 == 4245 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 18 Train mean loss: 436.99898854\n Train Accuracy%: 92.502 == 46251 / 50000\n Validation Accuracy%: 84.16 == 4208 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 19 Train mean loss: 416.98351576\n Train Accuracy%: 93.054 == 46527 / 50000\n Validation Accuracy%: 81.48 == 4074 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 20 Train mean loss: 346.36397807\n Train Accuracy%: 93.966 == 46983 / 50000\n Validation Accuracy%: 84.64 == 4232 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 21 Train mean loss: 296.34467645\n Train Accuracy%: 95.288 == 47644 / 50000\n Validation Accuracy%: 83.92 == 4196 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 22 Train mean loss: 310.18968987\n Train Accuracy%: 94.836 == 47418 / 50000\n Validation Accuracy%: 82.22 == 4111 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 23 Train mean loss: 544.12305131\n Train Accuracy%: 90.344 == 45172 / 50000\n Validation Accuracy%: 81.48 == 4074 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 24 Train mean loss: 507.90272054\n Train Accuracy%: 91.288 == 45644 / 50000\n Validation Accuracy%: 84.06 == 4203 / 5000\n-------------------------------------------------\n-------------------------------------------------\nEpoch: 25 Train mean loss: 377.94273479\n Train Accuracy%: 93.646 == 46823 / 50000\n Validation Accuracy%: 80.68 == 4034 / 5000\n-------------------------------------------------\n" ], [ "plt.plot(loss_hist[\"train accuracy\"])\nplt.plot(loss_hist[\"val accuracy\"])\nplt.xlabel(\"Epoch\")\nplt.ylabel(\"Loss\")\nplt.show()", "_____no_output_____" ], [ "plt.plot(loss_hist[\"train loss\"])\nplt.xlabel(\"Epoch\")\nplt.ylabel(\"Loss\")\nplt.show()", "_____no_output_____" ] ], [ [ "## Testing", "_____no_output_____" ] ], [ [ "with torch.no_grad():\n model.eval()\n \n y_true_test = []\n y_pred_test = []\n \n for batch_idx, (img, labels) in enumerate(test_loader):\n img = img.to(DEVICE)\n label = label.to(DEVICE)\n \n preds = model(img)\n \n y_pred_test.extend(preds.detach().argmax(dim=-1).tolist())\n y_true_test.extend(labels.detach().tolist())\n \n total_correct = len([True for x, y in zip(y_pred_test, y_true_test) if x==y])\n total = len(y_pred_test)\n accuracy = total_correct * 100 / total\n \n print(\"Test Accuracy%: \", accuracy, \"==\", total_correct, \"/\", total)", "Test Accuracy%: 81.04 == 4052 / 5000\n" ] ], [ [ "## Saving Model Weights ", "_____no_output_____" ] ], [ [ "torch.save(model.state_dict(), \"./trained_models/vgg16_cifar10.pt\")", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "pip list", "Package Version \n-------------------- ----------\nadal 1.2.2 \naltair 4.1.0 \nappdirs 1.4.3 \nappnope 0.1.0 \nattrs 19.3.0 \nazure-common 1.1.25 \nazure-kusto-data 0.0.44 \nazure-kusto-ingest 0.0.44 \nazure-storage-blob 2.1.0 \nazure-storage-common 2.1.0 \nazure-storage-queue 2.1.0 \nbackcall 0.1.0 \nbeakerx 1.4.1 \nblack 19.10b0 \nbleach 3.1.4 \nbqplot 0.12.6 \nbranca 0.3.1 \ncertifi 2020.4.5.1\ncffi 1.14.0 \nchardet 3.0.4 \nclick 7.1.2 \ncryptography 2.9.2 \ncycler 0.10.0 \ndebugpy 1.0.0b7 \ndecorator 4.4.2 \ndefusedxml 0.6.0 \nentrypoints 0.3 \nidna 2.9 \nipydatawidgets 4.0.1 \nipykernel 5.2.1 \nipyleaflet 0.12.4 \nipython 7.13.0 \nipython-genutils 0.2.0 \nipyvolume 0.5.2 \nipywebrtc 0.5.0 \nipywidgets 7.5.1 \nisodate 0.6.0 \njedi 0.17.0 \nJinja2 2.11.2 \njson5 0.9.5 \njsonschema 3.2.0 \njupyter-client 6.1.3 \njupyter-core 4.6.3 \njupyterlab 2.1.3 \njupyterlab-server 1.1.5 \nK3D 2.7.4 \nkiwisolver 1.2.0 \nMarkupSafe 1.1.1 \nmatplotlib 3.2.1 \nmistune 0.8.4 \nmsrest 0.6.13 \nmsrestazure 0.6.3 \nnbconvert 5.6.1 \nnbformat 5.0.5 \nnglview 2.7.5 \nnotebook 6.0.3 \nnumpy 1.18.2 \noauthlib 3.1.0 \npandas 1.0.3 \npandocfilters 1.4.2 \nparso 0.7.0 \npathspec 0.8.0 \npexpect 4.8.0 \npickleshare 0.7.5 \nPillow 7.1.1 \npip 19.2.3 \nprometheus-client 0.7.1 \nprompt-toolkit 3.0.5 \nptyprocess 0.6.0 \npy4j 0.10.9 \npycparser 2.20 \nPygments 2.6.1 \nPyJWT 1.7.1 \npyparsing 2.4.7 \npyrsistent 0.16.0 \npython-dateutil 2.8.1 \npythreejs 2.2.0 \npytz 2019.3 \npyzmq 19.0.0 \nqgrid 1.1.1 \nregex 2020.4.4 \nrequests 2.23.0 \nrequests-oauthlib 1.3.0 \nSend2Trash 1.5.0 \nsetuptools 41.2.0 \nsix 1.14.0 \nterminado 0.8.3 \ntestpath 0.4.4 \ntoml 0.10.0 \ntoolz 0.10.0 \ntornado 6.0.4 \ntraitlets 4.3.3 \ntraittypes 0.2.1 \ntyped-ast 1.4.1 \nurllib3 1.25.9 \nvega-datasets 0.8.0 \nwcwidth 0.1.9 \nwebencodings 0.5.1 \nwidgetsnbextension 3.5.1 \nxarray 0.15.1 \nNote: you may need to restart the kernel to use updated packages.\n" ] ], [ [ "# HELLO WORLD", "_____no_output_____" ] ], [ [ "with Error", "_____no_output_____" ], [ "import matplotlib\nimport matplotlib.pyplot as plt\nimport numpy as np\n\nt = np.arange(0.0, 2.0, 0.01)\ns = 1 + np.sin(2 * np.pi * t)\n\nfig, ax = plt.subplots()\nax.plot(t, s)\n\nax.set(xlabel='time (s)', ylabel='voltage (mV)',\n title='About as simple as it gets, folks')\n\nax.grid()\n\nfig.savefig('test.png')\n\nplt.show()", "_____no_output_____" ] ] ]

[ "code", "markdown", "code" ]

[ [ "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# _Mini Program - Working with SQLLite DB using Python_\n\n### <font color=green>Objective -</font>\n<font color=blue>1. This program gives an idea how to connect with SQLLite DB using Python and perform data manipulation </font><br>\n\n<font color=blue>2. There are 2 ways in which tables are create below to help you understand the robustness of this language</font>", "_____no_output_____" ], [ "### <font color=green>Step 1 - Import required libraries</font>\n#### <font color=blue>In this program we make used of 3 libraries</font>\n#### <font color=blue>1. sqlite3 - This module help to work with sql interface. It helps in performing db operations in sqllite database</font>\n#### <font color=blue>2. pandas - This module provides high performance and easy to use data manipulation and data analysis functionalities</font>\n#### <font color=blue>3. os - This module provides function to interact with operating system with easy use</font>\n \n", "_____no_output_____" ] ], [ [ "#Importing the required modules\nimport sqlite3\nimport pandas as pd\nimport os", "_____no_output_____" ] ], [ [ "### <font color=green>Step 2 - Creating a function to drop the table</font>\n#### <font color=blue>Function helps to re-create a reusable component that can be used conviniently and easily in other part of the code</font>\n#### <font color=blue>In Line 1 - We state the function name and specify the parameter being passed. In this case, the parameter is the table name</font>\n#### <font color=blue>In Line 2 - We write the sql query to be executed</font>\n#### <font color=blue>In Line 3 - We execute the query using the cursor object</font>", "_____no_output_____" ] ], [ [ "#Creating a function to drop the table if it exists\ndef dropTbl(tablename):\n dropTblStmt = \"DROP TABLE IF EXISTS \" + tablename\n c.execute(dropTblStmt)", "_____no_output_____" ] ], [ [ "### <font color=green>Step 3 - We create the database in which our table will reside</font>\n#### <font color=blue>In Line 1 - We are removing the already existing database file</font>\n#### <font color=blue>In Line 2 - We use connect function from the sqlite3 module to create a database studentGrades.db and establish a connection</font>\n#### <font color=blue>In Line 3 - We create a context of the database connection. This help to run all the database queries</font>", "_____no_output_____" ] ], [ [ "#Removing the database file\nos.remove('studentGrades.db')\n\n#Creating a new database - studentGrades.db\nconn = sqlite3.connect(\"studentGrades.db\")\nc = conn.cursor()", "_____no_output_____" ] ], [ [ "### <font color=green>Step 4 - We create a table in sqllite DB using data defined in the excel file</font>\n#### <font color=blue>This is the first method in which you can create a table. You can use to_sql function directly to read a dataframe and dump all it's content to the table</font>\n#### <font color=blue>In Line 1 - We are making use of dropTbl function created above to drop the table</font>\n#### <font color=blue>In Line 2 - We are creating a dataframe from the data read from the csv</font>\n#### <font color=blue>In Line 3 - We use to_sql function to push the data into the table. The first row of the file becomes the column name of the tables</font>\n#### <font color=blue>We repeat the above steps for all the 3 files to create 3 tables - STUDENT, GRADES and SUBJECTS</font>", "_____no_output_____" ] ], [ [ "#Reading data from csv file - student details, grades and subject\ndropTbl('STUDENT')\nstudent_details = pd.read_csv(\"Datafiles/studentDetails.csv\")\nstudent_details.to_sql('STUDENT',conn,index = False)\n\ndropTbl('GRADES')\nstudent_grades = pd.read_csv('Datafiles/studentGrades.csv')\nstudent_grades.to_sql('GRADES',conn,index = False)\n\ndropTbl('SUBJECTS')\nsubjects = pd.read_csv(\"Datafiles/subjects.csv\")\nsubjects.to_sql('SUBJECTS',conn,index = False)", "_____no_output_____" ] ], [ [ "### <font color=green>Step 5 - We create a master table STUDENT_GRADE_MASTER where we can colate the data from the individual tables by performing the joining operations</font>\n#### <font color=blue>In Line 1 - We are making use of dropTbl function created above to drop the table</font>\n#### <font color=blue>In Line 2 - We are writing sql query for table creation</font>\n#### <font color=blue>In Line 3 - We are using the cursor created above to execute the sql statement</font>\n#### <font color=blue>In Line 4 - We are using the second method of inserting data into the table. We are writing a query to insert the data after joining the data from all the tables</font>\n#### <font color=blue>In Line 5 - We are using the cursor created above to execute the sql statement</font>\n#### <font color=blue>In Line 6 - We are doing a commit operation. Since INSERT operation is a ddl, we have to perform a commit operation to register it into the database</font>", "_____no_output_____" ] ], [ [ "#Creating a table to store student master data\ndropTbl('STUDENT_GRADE_MASTER')\ncreateTblStmt = '''CREATE TABLE STUDENT_GRADE_MASTER\n ([Roll_number] INTEGER,\n [Student_Name] TEXT,\n [Stream] TEXT,\n [Subject] TEXT,\n [Marks] INTEGER\n )'''\nc.execute(createTblStmt)\n\n#Inserting data into the master table by joining the tables mentioned above\nqueryMaster = '''INSERT INTO STUDENT_GRADE_MASTER(Roll_number,Student_Name,Stream,Subject,Marks)\n SELECT g.roll_number, s.student_name, stream, sub.subject, marks from GRADES g \n LEFT OUTER JOIN STUDENT s on g.roll_number = s.roll_number\n LEFT OUTER JOIN SUBJECTS sub on g.subject_code = sub.subject_code'''\n\nc.execute(queryMaster)\nc.execute(\"COMMIT\")", "_____no_output_____" ] ], [ [ "### <font color=green>Step 6 - We can perform data fetch like we do in sqls using this sqlite3 module</font>\n#### <font color=blue>In Line 1 - We are writing a query to find the number of records in the master table</font>\n#### <font color=blue>In Line 2 - We are executing the above created query</font>\n#### <font color=blue>In Line 3 - fetchall function is used to get the result returned by the query. The result will be in the form of a list of tuples</font>\n#### <font color=blue>In Line 4 - We are writing another query to find the maximum marks recorded for each subject</font>\n#### <font color=blue>In Line 5 - We are executing the above created query</font>\n#### <font color=blue>In Line 6 - fetchall function is used to get the result returned by the query. The result will be in the form of a list of tuples</font>\n#### <font color=blue>In Line 7 - We are writing another query to find the percentage of marks obtained by each student in the class</font>\n#### <font color=blue>In Line 8 - We are executing the above created query</font>\n#### <font color=blue>In Line 9 - fetchall function is used to get the result returned by the query. The result will be in the form of a list of tuples</font>", "_____no_output_____" ] ], [ [ "#Finding the key data from the master table\n\n#1. Find the number of records in the master table\nquery_count = '''SELECT COUNT(*) FROM STUDENT_GRADE_MASTER'''\nc.execute(query_count)\nnumber_of_records = c.fetchall()\nprint(number_of_records)\n\n#2. Maximum marks for each subject\nquery_max_marks = '''SELECT Subject,max(Marks) as 'Max_Marks' from STUDENT_GRADE_MASTER GROUP BY Subject'''\nc.execute(query_max_marks)\nmax_marks_data = c.fetchall()\nprint(max_marks_data)\n\n#3. Percenatge of marks scored by each student\nquery_percentage = '''SELECT Student_Name, avg(Marks) as 'Percentage' from STUDENT_GRADE_MASTER GROUP BY Student_Name'''\nc.execute(query_percentage)\npercentage_data = c.fetchall()\nprint(percentage_data)\n", "[(20,)]\n[('C', 97), ('C++', 95), ('Environmental studies', 92), ('Java', 96), ('Maths', 98)]\n[('Abhishek', 94.2), ('Anand', 85.2), ('Sourabh', 89.0), ('Vivek', 84.8)]\n" ] ], [ [ "### <font color=green>Step 7 - We are closing the database connection</font>\n#### <font color=blue>It is always a good practice to close the database connection after all the operations are completed</font>", "_____no_output_____" ] ], [ [ "#Closing the connection\nconn.close()", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/AngieCat26/MujeresDigitales/blob/main/TALLER1.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ], [ "**PUNTO 2 **", "_____no_output_____" ] ], [ [ "premio1 = \"Viaje todo incluído para dos personas a San Andrés\"\npremio2 = \"una pasadía a los termales de San Vicente incluyendo almuerzo\"\npremio3 = \"Viaje todo incluido para dos personas a Santa Marta\"\npremio4 = \"Pasadía al desierto de Tatacoa (Sin incluír alimentación)\"\n\n\nrosada = premio1\nverde = premio2\nazul = premio3\ngris = premio4\nroja = \"No hay premio\"\n\n\ncliente = input(\"por favor digite el nombre del concursante \")\nbalota = input(\"Digite el color de la balota \")\nvalorVariable = int(input(\"Digite un valor variable \"))\n\n\nantiguedad = int(input(\"Digite los años de antiguedad del cliente \"))\nreferidos = int(input(\"Digite los referidos del cliente \"))\nliderazgo = input(\"¿El cliente tiene liderazgo en los programas de cooperación de viajes internos? \")\n\n\n\nif balota == \"rosada\":\n print(\"La empresa VIVAFLY se complace en anunciar que La participante \",cliente,\" ganó un \", rosada, \" en nuestro sorteo de viajes de AMOR y AMISTAD\")\n valorVariable1 = valorVariable * 0.15\n if valorVariable < 120000:\n print(\"El cliente \",cliente,\" también recibirá dos boletas de cine 4D con un combo de palomitas\")\n else :\n print(\"El cliente \", cliente, \" también recibirá un bono de $\",valorVariable1)\n\n \nelif balota == \"verde\":\n print(\"La empresa VIVAFLY se complace en anunciar que La participante \",cliente,\" ganó \", verde, \" en nuestro sorteo de viajes de AMOR y AMISTAD\")\n valorVariable2 = valorVariable * 0.20\n if valorVariable < 120000:\n print(\"El cliente \",cliente,\" también recibirá dos boletas de cine 4D con un combo de palomitas\")\n else :\n print(\"El cliente \", cliente, \" también recibirá un bono de $\",valorVariable2)\n\n\nelif balota == \"azul\":\n print(\"La empresa VIVAFLY se complace en anunciar que La participante \",cliente,\" ganó \", azul, \" en nuestro sorteo de viajes de AMOR y AMISTAD\")\n valorVariable3 = valorVariable * 0.05\n if valorVariable < 120000:\n print(\"El cliente \",cliente,\" también recibirá dos boletas de cine 4D con un combo de palomitas\")\n else :\n print(\"El cliente \", cliente, \" también recibirá un bono de $\",valorVariable3)\n\n\nelif balota == \"gris\":\n print(\"La empresa VIVAFLY se complace en anunciar que La participante \",cliente,\" ganó \", gris, \" en nuestro sorteo de viajes de AMOR y AMISTAD\")\n valorVariable4 = valorVariable * 0.20\n if valorVariable < 120000:\n print(\"El cliente \",cliente,\" también recibirá dos boletas de cine 4D con un combo de palomitas\")\n else :\n print(\"El cliente \", cliente, \" también recibirá un bono de $\",valorVariable4)\n\nelse :\n print(\"La empresa VIVAFLY se complace en anunciar que La participante \",cliente,\" ganó $120.000\")\n\n\n\n#si se cumple los 3:\n\nif antiguedad > 10 and referidos > 10 and liderazgo == \"si\" :\n puntos = valorVariable * 0.50\n print(\"Adicionalmente:\")\n if puntos > 600000 : \n puntos = 500 * 0.20 \n print(\"Se le otorga \",puntos, \" puntos para redimir en un premio a futuro\")\n else :\n puntos = 250 * 0.20 \n print(\"Se le otorga \", puntos, \" puntos para redimir en un premio a futuro\")\n\n#Si se cumple solo dos requisitos:\n\nelif antiguedad > 10 and referidos > 10 :\n puntos = valorVariable * 0.50\n print(\"Adicionalmente: \")\n if puntos > 250000 :\n puntos = 200 * 0.20 \n print(\"Se le otorga \", puntos, \" puntos para redimir en un premio a futuro\")\n else :\n puntos = 50 * 0.20 \n print(\"Se le otorga \", puntos, \" puntos para redimir en un premio a futuro\")\n\nelif antiguedad > 10 and liderazgo == \"si\" :\n puntos = valorVariable * 0.50\n print(\"Adicionalmente: \")\n if puntos > 250000 : \n puntos = 200 * 0.20 \n print(\"Se le otorga \", puntos, \" puntos para redimir en un premio a futuro\")\n else :\n puntos = 50 * 0.20 \n print(\"Se le otorga \", puntos, \"puntos para redimir en un premio a futuro\")\n\nelif referidos > 10 and liderazgo ==\"si\" :\n puntos = valorVariable * 0.30\n print(\"Adicionalmente: \")\n if puntos > 250000 : \n puntos = 200 * 0.20 \n print(\"Se le otorga 200 puntos para redimir en un premio a futuro\")\n else :\n puntos = 50 * 0.20 \n print(\"Se le otorga 50 puntos para redimir en un premio a futuro\")\n\n\n#Solo si se cumple uno\nelse : \n puntos = valorVariable * 0.10\n puntos = puntos * 0.10\n print(\"Adicionalmente se se le otorga\", puntos, \"puntos para redimir en un premio a futuro\")\n", "por favor digite el nombre del concursante Angie\nDigite el color de la balota rosada\nDigite un valor variable 2000000\nDigite los años de antiguedad del cliente 14\nDigite los referidos del cliente 1\n¿El cliente tiene liderazgo en los programas de cooperación de viajes internos? si\nLa empresa VIVAFLY se complace en anunciar que La participante Angie ganó un Viaje todo incluído para dos personas a San Andrés en nuestro sorteo de viajes de AMOR y AMISTAD\nEl cliente Angie también recibirá un bono de $ 300000.0\nAdicionalmente: \nSe le otorga 40.0 puntos para redimir en un premio a futuro\n" ] ] ]

[ "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Tutorial\n#### This tutorial will introduce you to the *fifa_preprocessing*'s functionality!\nIn general, the following functions will alow you to preprocess your data to be able to perform machine learning or statistical data analysis by reformatting, casting or deleting certain values.", "_____no_output_____" ], [ "The data used in these examples comes from https://www.kaggle.com/karangadiya/fifa19, a webpage this package was inspired by. The module's functions work best with this data set, however, they will work with any data structured in a similar manner.", "_____no_output_____" ], [ "## Prerequisites", "_____no_output_____" ], [ "First, import the fifa_preprocessing, pandas and math modules:", "_____no_output_____" ] ], [ [ "import fifa_preprocessing as fp\nimport pandas as pd\nimport math", "_____no_output_____" ] ], [ [ "Load your data:", "_____no_output_____" ] ], [ [ "data = pd.read_csv('data.csv')", "_____no_output_____" ], [ "data", "_____no_output_____" ] ], [ [ "## Exclude goalkeepers\nBefore any preprocessing, the data contains all the players.", "_____no_output_____" ] ], [ [ "data[['Name', 'Position']]", "_____no_output_____" ] ], [ [ "This command will exclude goalkeepers from your data set (i.e. delete all the rows where column 'Position' is equal to 'GK'):", "_____no_output_____" ] ], [ [ "data = fp.exclude_goalkeepers(data)", "_____no_output_____" ] ], [ [ "As you may notice, the row number 3 was deleted.", "_____no_output_____" ] ], [ [ "data[['Name', 'Position']]", "_____no_output_____" ] ], [ [ "## Format currencies\nTo remove unnecessary characters form a monetary value use:", "_____no_output_____" ] ], [ [ "money = '€23.4M'\nfp.money_format(money)", "_____no_output_____" ] ], [ [ "The value will be expressed in thousands of euros:", "_____no_output_____" ] ], [ [ "money = '€7K'\nfp.money_format(money)", "_____no_output_____" ] ], [ [ "## Format players' rating\nIn FIFA players get a ranking on they skills on the pitch. The ranking is represented as a sum of two integers.\nThe following function lets you take in a string containing two numbers separated by a '+' and get the actual sum:", "_____no_output_____" ] ], [ [ "rating = '81+3'\nfp.rating_format(rating)", "_____no_output_____" ] ], [ [ "## Format players' work rate\nThe next function takes in a qualitative parameter that could be expressed as a quantitive value.\nIf you have a data set where one category is expressed as 'High', 'Medium' or 'Low', this function will assign numbers to these values (2, 1 and 0 respectively):", "_____no_output_____" ] ], [ [ "fp.work_format('High')", "_____no_output_____" ], [ "fp.work_format('Medium')", "_____no_output_____" ], [ "fp.work_format('Low')", "_____no_output_____" ] ], [ [ "In fact, the function returns 0 in every case where the passed in parameter id different than 'High' and 'Medium':\n", "_____no_output_____" ] ], [ [ "fp.work_format('Mediocre')", "_____no_output_____" ] ], [ [ "## Cast to int\nThis simple function casts a float to int, but also adds extra flexibility and returns 0 when it encounters a NaN (Not a Number):", "_____no_output_____" ] ], [ [ "fp.to_int(3.24)", "_____no_output_____" ], [ "import numpy\nnan = numpy.nan\nfp.to_int(nan)", "_____no_output_____" ] ], [ [ "## Apply format of choice\nThis generic function lets you choose what format to apply to every value in the columns of the data frame you specify.", "_____no_output_____" ] ], [ [ "data[['Name', 'Jersey Number', 'Skill Moves', 'Weak Foot']]", "_____no_output_____" ] ], [ [ "By format, is meant a function that operates on the values in the specified columns:", "_____no_output_____" ] ], [ [ "columns = ['Jersey Number', 'Skill Moves', 'Weak Foot']\nformat_fun = fp.to_int\n\ndata = fp.apply_format(data, columns, format_fun)\n\ndata[['Name', 'Jersey Number', 'Skill Moves', 'Weak Foot']]", "_____no_output_____" ] ], [ [ "## Dummy variables\nIf we intend to build machine learning models to explore our data, we usually are not able to extract any information from qualitative data. Here 'Club' and 'Preferred Foot' are categories that could bring interesting information. To be able to use it in our machine learning algorithms we can get dummy variables.", "_____no_output_____" ] ], [ [ "data[['Name', 'Preferred Foot']]", "_____no_output_____" ] ], [ [ "If we choose 'Preferred Foot', new columns will be aded, their titles will be the same as the values in 'Preferred Foot' column: 'Left' and 'Right'. So now instead of seeing 'Left' in the column 'Preferred Foot' we will see 1 in 'Left' column (and 0 in 'Right').", "_____no_output_____" ] ], [ [ "data = fp.to_dummy(data, ['Preferred Foot'])\ndata[['Name', 'Left', 'Right']]", "_____no_output_____" ] ], [ [ "Learn more about [dummy variables](https://en.wikiversity.org/wiki/Dummy_variable_(statistics)).", "_____no_output_____" ], [ "The data frame will no longer contain the columns we transformed:", "_____no_output_____" ] ], [ [ "'Preferred Foot' in data", "_____no_output_____" ] ], [ [ "We can get dummy variables for multiple columns at once.", "_____no_output_____" ] ], [ [ "data[['Name', 'Club', 'Position']]", "_____no_output_____" ], [ "data = fp.to_dummy(data, ['Club', 'Nationality'])\ndata[['Name', 'Paris Saint-Germain', 'Manchester City', 'Brazil', 'Portugal']]", "_____no_output_____" ] ], [ [ "## Split work rate column\nIn FIFA the players' work rate is saved in a special way, two qualiative values are split with a slash:", "_____no_output_____" ] ], [ [ "data[['Name', 'Work Rate']]", "_____no_output_____" ] ], [ [ "This next function allows you to split column 'Work Rate' into 'Defensive Work Rate' and 'Offensive Work Rate':", "_____no_output_____" ] ], [ [ "data = fp.split_work_rate(data)\ndata[['Name', 'Defensive Work Rate', 'Offensive Work Rate']]", "_____no_output_____" ] ], [ [ "## Default preprocessing\nTo perform all the basic preprocessing (optimized for the FIFA 19 data set) on your data, simply go:", "_____no_output_____" ] ], [ [ "data = pd.read_csv('data.csv')\nfp.preprocess(data)", "_____no_output_____" ] ], [ [ "## Let's get coding\nNow it is your turn to try out our functions and preprocess your data!", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Bias", "_____no_output_____" ], [ "### Goals\nIn this notebook, you're going to explore a way to identify some biases of a GAN using a classifier, in a way that's well-suited for attempting to make a model independent of an input. Note that not all biases are as obvious as the ones you will see here.\n\n### Learning Objectives\n1. Be able to distinguish a few different kinds of bias in terms of demographic parity, equality of odds, and equality of opportunity (as proposed [here](http://m-mitchell.com/papers/Adversarial_Bias_Mitigation.pdf)).\n2. Be able to use a classifier to try and detect biases in a GAN by analyzing the generator's implicit associations.", "_____no_output_____" ], [ "\n## Challenges\n\nOne major challenge in assessing bias in GANs is that you still want your generator to be able to generate examples of different values of a protected class—the class you would like to mitigate bias against. While a classifier can be optimized to have its output be independent of a protected class, a generator which generates faces should be able to generate examples of various protected class values. \n\nWhen you generate examples with various values of a protected class, you don’t want those examples to correspond to any properties that aren’t strictly a function of that protected class. This is made especially difficult since many protected classes (e.g. gender or ethnicity) are social constructs, and what properties count as “a function of that protected class” will vary depending on who you ask. It’s certainly a hard balance to strike.\n\nMoreover, a protected class is rarely used to condition a GAN explicitly, so it is often necessary to resort to somewhat post-hoc methods (e.g. using a classifier trained on relevant features, which might be biased itself). \n\nIn this assignment, you will learn one approach to detect potential bias, by analyzing correlations in feature classifications on the generated images. ", "_____no_output_____" ], [ "## Getting Started\n\nAs you have done previously, you will start by importing some useful libraries and defining a visualization function for your images. You will also use the same generator and basic classifier from previous weeks.", "_____no_output_____" ], [ "#### Packages and Visualization", "_____no_output_____" ] ], [ [ "import torch\nimport numpy as np\nfrom torch import nn\nfrom tqdm.auto import tqdm\nfrom torchvision import transforms\nfrom torchvision.utils import make_grid\nfrom torchvision.datasets import CelebA\nfrom torch.utils.data import DataLoader\nimport matplotlib.pyplot as plt\ntorch.manual_seed(0) # Set for our testing purposes, please do not change!\n\ndef show_tensor_images(image_tensor, num_images=16, size=(3, 64, 64), nrow=3):\n '''\n Function for visualizing images: Given a tensor of images, number of images,\n size per image, and images per row, plots and prints the images in an uniform grid.\n '''\n image_tensor = (image_tensor + 1) / 2\n image_unflat = image_tensor.detach().cpu()\n image_grid = make_grid(image_unflat[:num_images], nrow=nrow)\n plt.imshow(image_grid.permute(1, 2, 0).squeeze())\n plt.show()", "_____no_output_____" ] ], [ [ "#### Generator and Noise", "_____no_output_____" ] ], [ [ "class Generator(nn.Module):\n '''\n Generator Class\n Values:\n z_dim: the dimension of the noise vector, a scalar\n im_chan: the number of channels in the images, fitted for the dataset used, a scalar\n (CelebA is rgb, so 3 is your default)\n hidden_dim: the inner dimension, a scalar\n '''\n def __init__(self, z_dim=10, im_chan=3, hidden_dim=64):\n super(Generator, self).__init__()\n self.z_dim = z_dim\n # Build the neural network\n self.gen = nn.Sequential(\n self.make_gen_block(z_dim, hidden_dim * 8),\n self.make_gen_block(hidden_dim * 8, hidden_dim * 4),\n self.make_gen_block(hidden_dim * 4, hidden_dim * 2),\n self.make_gen_block(hidden_dim * 2, hidden_dim),\n self.make_gen_block(hidden_dim, im_chan, kernel_size=4, final_layer=True),\n )\n\n def make_gen_block(self, input_channels, output_channels, kernel_size=3, stride=2, final_layer=False):\n '''\n Function to return a sequence of operations corresponding to a generator block of DCGAN;\n a transposed convolution, a batchnorm (except in the final layer), and an activation.\n Parameters:\n input_channels: how many channels the input feature representation has\n output_channels: how many channels the output feature representation should have\n kernel_size: the size of each convolutional filter, equivalent to (kernel_size, kernel_size)\n stride: the stride of the convolution\n final_layer: a boolean, true if it is the final layer and false otherwise \n (affects activation and batchnorm)\n '''\n if not final_layer:\n return nn.Sequential(\n nn.ConvTranspose2d(input_channels, output_channels, kernel_size, stride),\n nn.BatchNorm2d(output_channels),\n nn.ReLU(inplace=True),\n )\n else:\n return nn.Sequential(\n nn.ConvTranspose2d(input_channels, output_channels, kernel_size, stride),\n nn.Tanh(),\n )\n\n def forward(self, noise):\n '''\n Function for completing a forward pass of the generator: Given a noise tensor, \n returns generated images.\n Parameters:\n noise: a noise tensor with dimensions (n_samples, z_dim)\n '''\n x = noise.view(len(noise), self.z_dim, 1, 1)\n return self.gen(x)\n\ndef get_noise(n_samples, z_dim, device='cpu'):\n '''\n Function for creating noise vectors: Given the dimensions (n_samples, z_dim)\n creates a tensor of that shape filled with random numbers from the normal distribution.\n Parameters:\n n_samples: the number of samples to generate, a scalar\n z_dim: the dimension of the noise vector, a scalar\n device: the device type\n '''\n return torch.randn(n_samples, z_dim, device=device)", "_____no_output_____" ] ], [ [ "#### Classifier", "_____no_output_____" ] ], [ [ "class Classifier(nn.Module):\n '''\n Classifier Class\n Values:\n im_chan: the number of channels in the images, fitted for the dataset used, a scalar\n (CelebA is rgb, so 3 is your default)\n n_classes: the total number of classes in the dataset, an integer scalar\n hidden_dim: the inner dimension, a scalar\n '''\n def __init__(self, im_chan=3, n_classes=2, hidden_dim=64):\n super(Classifier, self).__init__()\n self.classifier = nn.Sequential(\n self.make_classifier_block(im_chan, hidden_dim),\n self.make_classifier_block(hidden_dim, hidden_dim * 2),\n self.make_classifier_block(hidden_dim * 2, hidden_dim * 4, stride=3),\n self.make_classifier_block(hidden_dim * 4, n_classes, final_layer=True),\n )\n\n def make_classifier_block(self, input_channels, output_channels, kernel_size=4, stride=2, final_layer=False):\n '''\n Function to return a sequence of operations corresponding to a classifier block; \n a convolution, a batchnorm (except in the final layer), and an activation (except in the final layer).\n Parameters:\n input_channels: how many channels the input feature representation has\n output_channels: how many channels the output feature representation should have\n kernel_size: the size of each convolutional filter, equivalent to (kernel_size, kernel_size)\n stride: the stride of the convolution\n final_layer: a boolean, true if it is the final layer and false otherwise \n (affects activation and batchnorm)\n '''\n if not final_layer:\n return nn.Sequential(\n nn.Conv2d(input_channels, output_channels, kernel_size, stride),\n nn.BatchNorm2d(output_channels),\n nn.LeakyReLU(0.2, inplace=True),\n )\n else:\n return nn.Sequential(\n nn.Conv2d(input_channels, output_channels, kernel_size, stride),\n )\n\n def forward(self, image):\n '''\n Function for completing a forward pass of the classifier: Given an image tensor, \n returns an n_classes-dimension tensor representing classes.\n Parameters:\n image: a flattened image tensor with im_chan channels\n '''\n class_pred = self.classifier(image)\n return class_pred.view(len(class_pred), -1)", "_____no_output_____" ] ], [ [ "## Specifying Parameters\nYou will also need to specify a few parameters before you begin training:\n * z_dim: the dimension of the noise vector\n * batch_size: the number of images per forward/backward pass\n * device: the device type", "_____no_output_____" ] ], [ [ "z_dim = 64\nbatch_size = 128\ndevice = 'cuda'", "_____no_output_____" ] ], [ [ "## Train a Classifier (Optional)\n\nYou're welcome to train your own classifier with this code, but you are provide a pre-trained one based on this architecture here which you can load and use in the next section. ", "_____no_output_____" ] ], [ [ "# You can run this code to train your own classifier, but there is a provided pre-trained one \n# If you'd like to use this, just run \"train_classifier(filename)\"\n# To train and save a classifier on the label indices to that filename\ndef train_classifier(filename):\n import seaborn as sns\n import matplotlib.pyplot as plt\n\n # You're going to target all the classes, so that's how many the classifier will learn\n label_indices = range(40)\n\n n_epochs = 3\n display_step = 500\n lr = 0.001\n beta_1 = 0.5\n beta_2 = 0.999\n image_size = 64\n\n transform = transforms.Compose([\n transforms.Resize(image_size),\n transforms.CenterCrop(image_size),\n transforms.ToTensor(),\n transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),\n ])\n\n dataloader = DataLoader(\n CelebA(\".\", split='train', download=True, transform=transform),\n batch_size=batch_size,\n shuffle=True)\n\n classifier = Classifier(n_classes=len(label_indices)).to(device)\n class_opt = torch.optim.Adam(classifier.parameters(), lr=lr, betas=(beta_1, beta_2))\n criterion = nn.BCEWithLogitsLoss()\n\n cur_step = 0\n classifier_losses = []\n # classifier_val_losses = []\n for epoch in range(n_epochs):\n # Dataloader returns the batches\n for real, labels in tqdm(dataloader):\n real = real.to(device)\n labels = labels[:, label_indices].to(device).float()\n\n class_opt.zero_grad()\n class_pred = classifier(real)\n class_loss = criterion(class_pred, labels)\n class_loss.backward() # Calculate the gradients\n class_opt.step() # Update the weights\n classifier_losses += [class_loss.item()] # Keep track of the average classifier loss\n\n ### Visualization code ###\n if cur_step % display_step == 0 and cur_step > 0:\n class_mean = sum(classifier_losses[-display_step:]) / display_step\n print(f\"Step {cur_step}: Classifier loss: {class_mean}\")\n step_bins = 20\n x_axis = sorted([i * step_bins for i in range(len(classifier_losses) // step_bins)] * step_bins)\n sns.lineplot(x_axis, classifier_losses[:len(x_axis)], label=\"Classifier Loss\")\n plt.legend()\n plt.show()\n torch.save({\"classifier\": classifier.state_dict()}, filename)\n cur_step += 1\n\n# Uncomment the last line to train your own classfier - this line will not work in Coursera.\n# If you'd like to do this, you'll have to download it and run it, ideally using a GPU.\n# train_classifier(\"filename\")", "_____no_output_____" ] ], [ [ "## Loading the Pre-trained Models\n\nYou can now load the pre-trained generator (trained on CelebA) and classifier using the following code. If you trained your own classifier, you can load that one here instead. However, it is suggested that you first go through the assignment using the pre-trained one.", "_____no_output_____" ] ], [ [ "import torch\ngen = Generator(z_dim).to(device)\ngen_dict = torch.load(\"pretrained_celeba.pth\", map_location=torch.device(device))[\"gen\"]\ngen.load_state_dict(gen_dict)\ngen.eval()\n\nn_classes = 40\nclassifier = Classifier(n_classes=n_classes).to(device)\nclass_dict = torch.load(\"pretrained_classifier.pth\", map_location=torch.device(device))[\"classifier\"]\nclassifier.load_state_dict(class_dict)\nclassifier.eval()\nprint(\"Loaded the models!\")\n\nopt = torch.optim.Adam(classifier.parameters(), lr=0.01)", "_____no_output_____" ] ], [ [ "## Feature Correlation\nNow you can generate images using the generator. By also using the classifier, you will be generating images with different amounts of the \"male\" feature.\n\nYou are welcome to experiment with other features as the target feature, but it is encouraged that you initially go through the notebook as is before exploring.", "_____no_output_____" ] ], [ [ "# First you generate a bunch of fake images with the generator\nn_images = 256\nfake_image_history = []\nclassification_history = []\ngrad_steps = 30 # How many gradient steps to take\nskip = 2 # How many gradient steps to skip in the visualization\n\nfeature_names = [\"5oClockShadow\", \"ArchedEyebrows\", \"Attractive\", \"BagsUnderEyes\", \"Bald\", \"Bangs\",\n\"BigLips\", \"BigNose\", \"BlackHair\", \"BlondHair\", \"Blurry\", \"BrownHair\", \"BushyEyebrows\", \"Chubby\",\n\"DoubleChin\", \"Eyeglasses\", \"Goatee\", \"GrayHair\", \"HeavyMakeup\", \"HighCheekbones\", \"Male\", \n\"MouthSlightlyOpen\", \"Mustache\", \"NarrowEyes\", \"NoBeard\", \"OvalFace\", \"PaleSkin\", \"PointyNose\", \n\"RecedingHairline\", \"RosyCheeks\", \"Sideburn\", \"Smiling\", \"StraightHair\", \"WavyHair\", \"WearingEarrings\", \n\"WearingHat\", \"WearingLipstick\", \"WearingNecklace\", \"WearingNecktie\", \"Young\"]\n\nn_features = len(feature_names)\n# Set the target feature\ntarget_feature = \"Male\"\ntarget_indices = feature_names.index(target_feature)\nnoise = get_noise(n_images, z_dim).to(device)\nnew_noise = noise.clone().requires_grad_()\nstarting_classifications = classifier(gen(new_noise)).cpu().detach()\n\n# Additive direction (more of a feature)\nfor i in range(grad_steps):\n opt.zero_grad()\n fake = gen(new_noise)\n fake_image_history += [fake]\n classifications = classifier(fake)\n classification_history += [classifications.cpu().detach()]\n fake_classes = classifications[:, target_indices].mean()\n fake_classes.backward()\n new_noise.data += new_noise.grad / grad_steps\n\n# Subtractive direction (less of a feature)\nnew_noise = noise.clone().requires_grad_()\nfor i in range(grad_steps):\n opt.zero_grad()\n fake = gen(new_noise)\n fake_image_history += [fake]\n classifications = classifier(fake)\n classification_history += [classifications.cpu().detach()]\n fake_classes = classifications[:, target_indices].mean()\n fake_classes.backward()\n new_noise.data -= new_noise.grad / grad_steps\n\nclassification_history = torch.stack(classification_history)", "_____no_output_____" ], [ "print(classification_history.shape)\nprint(starting_classifications[None, :, :].shape)", "_____no_output_____" ] ], [ [ "You've now generated image samples, which have increasing or decreasing amounts of the target feature. You can visualize the way in which that affects other classified features. The x-axis will show you the amount of change in your target feature and the y-axis shows how much the other features change, as detected in those images by the classifier. Together, you will be able to see the covariance of \"male-ness\" and other features.\n\nYou are started off with a set of features that have interesting associations with \"male-ness\", but you are welcome to change the features in `other_features` with others from `feature_names`.", "_____no_output_____" ] ], [ [ "import seaborn as sns\n# Set the other features\nother_features = [\"Smiling\", \"Bald\", \"Young\", \"HeavyMakeup\", \"Attractive\"]\nclassification_changes = (classification_history - starting_classifications[None, :, :]).numpy()\nfor other_feature in other_features:\n other_indices = feature_names.index(other_feature)\n with sns.axes_style(\"darkgrid\"):\n sns.regplot(\n classification_changes[:, :, target_indices].reshape(-1), \n classification_changes[:, :, other_indices].reshape(-1), \n fit_reg=True,\n truncate=True,\n ci=99,\n x_ci=99,\n x_bins=len(classification_history),\n label=other_feature\n )\nplt.xlabel(target_feature)\nplt.ylabel(\"Other Feature\")\nplt.title(f\"Generator Biases: Features vs {target_feature}-ness\")\nplt.legend(loc=1)\nplt.show()", "_____no_output_____" ] ], [ [ "This correlation detection can be used to reduce bias by penalizing this type of correlation in the loss during the training of the generator. However, currently there is no rigorous and accepted solution for debiasing GANs. A first step that you can take in the right direction comes before training the model: make sure that your dataset is inclusive and representative, and consider how you can mitigate the biases resulting from whatever data collection method you used—for example, getting a representative labelers for your task. \n\nIt is important to note that, as highlighted in the lecture and by many researchers including [Timnit Gebru and Emily Denton](https://sites.google.com/view/fatecv-tutorial/schedule), a diverse dataset alone is not enough to eliminate bias. Even diverse datasets can reinforce existing structural biases by simply capturing common social biases. Mitigating these biases is an important and active area of research.\n\n#### Note on CelebA\nYou may have noticed that there are obvious correlations between the feature you are using, \"male\", and other seemingly unrelates features, \"smiling\" and \"young\" for example. This is because the CelebA dataset labels had no serious consideration for diversity. The data represents the biases their labelers, the dataset creators, the social biases as a result of using a dataset based on American celebrities, and many others. Equipped with knowledge about bias, we trust that you will do better in the future datasets you create.", "_____no_output_____" ], [ "## Quantification\nFinally, you can also quantitatively evaluate the degree to which these factors covary. Given a target index, for example corresponding to \"male,\" you'll want to return the other features that covary with that target feature the most. You'll want to account for both large negative and positive covariances, and you'll want to avoid returning the target feature in your list of covarying features (since a feature will often have a high covariance with itself).\n\n<details>\n\n<summary>\n<font size=\"3\" color=\"green\">\n<b>Optional hints for <code><font size=\"4\">get_top_covariances</font></code></b>\n</font>\n</summary>\n\n1. You will likely find the following function useful: [np.cov](https://numpy.org/doc/stable/reference/generated/numpy.cov.html).\n2. You will probably find it useful to [reshape](https://numpy.org/doc/stable/reference/generated/numpy.reshape.html) the input.\n3. The target feature should not be included in the outputs.\n4. Feel free to use any reasonable method to get the top-n elements.\n5. It may be easiest to solve this if you find the `relevant_indices` first.\n6. You want to sort by absolute value but return the actual values.\n</details>", "_____no_output_____" ] ], [ [ "# UNQ_C1 (UNIQUE CELL IDENTIFIER, DO NOT EDIT)\n# GRADED CELL: get_top_covariances\ndef get_top_covariances(classification_changes, target_index, top_n=10):\n '''\n Function for getting the top n covariances: Given a list of classification changes\n and the index of the target feature, returns (1) a list or tensor (numpy or torch) of the indices\n corresponding to the n features that covary most with the target in terms of absolute covariance\n and (2) a list or tensor (numpy or torch) of the degrees to which they covary.\n Parameters:\n classification_changes: relative changes in classifications of each generated image \n resulting from optimizing the target feature (see above for a visualization)\n target_index: the index of the target feature, a scalar\n top_n: the top most number of elements to return, default is 10\n '''\n # Hint: Don't forget you also care about negative covariances!\n # Note that classification_changes has a shape of (2 * grad_steps, n_images, n_features) \n # where n_features is the number of features measured by the classifier, and you are looking\n # for the covariance of the features based on the (2 * grad_steps * n_images) samples\n #### START CODE HERE ####\n relevant_indices = None\n highest_covariances = None\n #### END CODE HERE ####\n return relevant_indices, highest_covariances", "_____no_output_____" ], [ "# UNIT TEST\nfrom torch.distributions import MultivariateNormal\nmean = torch.Tensor([0, 0, 0, 0]) \ncovariance = torch.Tensor( \n [[10, 2, -0.5, -5],\n [2, 11, 5, 4],\n [-0.5, 5, 10, 2],\n [-5, 4, 2, 11]]\n)\nindependent_dist = MultivariateNormal(mean, covariance)\nsamples = independent_dist.sample((60 * 128,))\nfoo = samples.reshape(60, 128, samples.shape[-1])\n\nrelevant_indices, highest_covariances = get_top_covariances(foo, 1, top_n=3)\nassert (tuple(relevant_indices) == (2, 3, 0)), \"Make sure you're getting the greatest, not the least covariances\"\nassert np.all(np.abs(highest_covariances - [5, 4, 2]) < 0.5 )\n\nrelevant_indices, highest_covariances = get_top_covariances(foo, 0, top_n=3)\nassert (tuple(relevant_indices) == (3, 1, 2)), \"Make sure to consider the magnitude of negative covariances\"\nassert np.all(np.abs(highest_covariances - [-5, 2, -0.5]) < 0.5 )\n\nrelevant_indices, highest_covariances = get_top_covariances(foo, 2, top_n=2)\nassert (tuple(relevant_indices) == (1, 3))\nassert np.all(np.abs(highest_covariances - [5, 2]) < 0.5 )\n\nrelevant_indices, highest_covariances = get_top_covariances(foo, 3, top_n=2)\nassert (tuple(relevant_indices) == (0, 1))\nassert np.all(np.abs(highest_covariances - [-5, 4]) < 0.5 )\n\nprint(\"All tests passed\")", "_____no_output_____" ], [ "relevant_indices, highest_covariances = get_top_covariances(classification_changes, target_indices, top_n=10)\nprint(relevant_indices)\nassert relevant_indices[9] == 34\nassert len(relevant_indices) == 10\nassert highest_covariances[8] - (-1.2418) < 1e-3\nfor index, covariance in zip(relevant_indices, highest_covariances):\n print(f\"{feature_names[index]} {covariance:f}\")", "_____no_output_____" ] ], [ [ "One of the major sources of difficulty with identifying bias and fairness, as discussed in the lectures, is that there are many ways you might reasonably define these terms. Here are three ways that are computationally useful and [widely referenced](http://m-mitchell.com/papers/Adversarial_Bias_Mitigation.pdf). They are, by no means, the only definitions of fairness (see more details [here](https://developers.google.com/machine-learning/glossary/fairness)):\n\n\n1. Demographic parity: the overall distribution of the predictions made by a predictor is the same for different values of a protected class. \n2. Equality of odds: all else being equal, the probability that you predict correctly or incorrectly is the same for different values of a protected class. \n2. Equality of opportunity: all else being equal, the probability that you predict correctly is the same for different valus of a protected class (weaker than equality of odds).\n\nWith GANs also being used to help downstream classifiers (you will see this firsthand in future assignments), these definitions of fairness will impact, as well as depend on, your downstream task. It is important to work towards creating a fair GAN according to the definition you choose. Pursuing any of them is virtually always better than blindly labelling data, creating a GAN, and sampling its generations.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Neural network hybrid recommendation system on Google Analytics data model and training\n\nThis notebook demonstrates how to implement a hybrid recommendation system using a neural network to combine content-based and collaborative filtering recommendation models using Google Analytics data. We are going to use the learned user embeddings from [wals.ipynb](../wals.ipynb) and combine that with our previous content-based features from [content_based_using_neural_networks.ipynb](../content_based_using_neural_networks.ipynb)\n\nNow that we have our data preprocessed from BigQuery and Cloud Dataflow, we can build our neural network hybrid recommendation model to our preprocessed data. Then we can train locally to make sure everything works and then use the power of Google Cloud ML Engine to scale it out.", "_____no_output_____" ], [ "We're going to use TensorFlow Hub to use trained text embeddings, so let's first pip install that and reset our session.", "_____no_output_____" ] ], [ [ "!pip install tensorflow_hub", "_____no_output_____" ] ], [ [ "Now reset the notebook's session kernel! Since we're no longer using Cloud Dataflow, we'll be using the python3 kernel from here on out so don't forget to change the kernel if it's still python2.", "_____no_output_____" ] ], [ [ "# Import helpful libraries and setup our project, bucket, and region\nimport os\nimport tensorflow as tf\nimport tensorflow_hub as hub\n\nPROJECT = 'cloud-training-demos' # REPLACE WITH YOUR PROJECT ID\nBUCKET = 'cloud-training-demos-ml' # REPLACE WITH YOUR BUCKET NAME\nREGION = 'us-central1' # REPLACE WITH YOUR BUCKET REGION e.g. us-central1\n\n# do not change these\nos.environ['PROJECT'] = PROJECT\nos.environ['BUCKET'] = BUCKET\nos.environ['REGION'] = REGION\nos.environ['TFVERSION'] = '1.8'", "_____no_output_____" ], [ "%%bash\ngcloud config set project $PROJECT\ngcloud config set compute/region $REGION", "_____no_output_____" ], [ "%%bash\nif ! gsutil ls | grep -q gs://${BUCKET}/hybrid_recommendation/preproc; then\n gsutil mb -l ${REGION} gs://${BUCKET}\n # copy canonical set of preprocessed files if you didn't do preprocessing notebook\n gsutil -m cp -R gs://cloud-training-demos/courses/machine_learning/deepdive/10_recommendation/hybrid_recommendation gs://${BUCKET}\nfi", "_____no_output_____" ] ], [ [ "<h2> Create hybrid recommendation system model using TensorFlow </h2>\n\nNow that we've created our training and evaluation input files as well as our categorical feature vocabulary files, we can create our TensorFlow hybrid recommendation system model.", "_____no_output_____" ], [ "Let's first get some of our aggregate information that we will use in the model from some of our preprocessed files we saved in Google Cloud Storage.", "_____no_output_____" ] ], [ [ "from tensorflow.python.lib.io import file_io", "_____no_output_____" ], [ "# Get number of content ids from text file in Google Cloud Storage\nwith file_io.FileIO(tf.gfile.Glob(filename = \"gs://{}/hybrid_recommendation/preproc/vocab_counts/content_id_vocab_count.txt*\".format(BUCKET))[0], mode = 'r') as ifp:\n number_of_content_ids = int([x for x in ifp][0])\nprint(\"number_of_content_ids = {}\".format(number_of_content_ids))", "_____no_output_____" ], [ "# Get number of categories from text file in Google Cloud Storage\nwith file_io.FileIO(tf.gfile.Glob(filename = \"gs://{}/hybrid_recommendation/preproc/vocab_counts/category_vocab_count.txt*\".format(BUCKET))[0], mode = 'r') as ifp:\n number_of_categories = int([x for x in ifp][0])\nprint(\"number_of_categories = {}\".format(number_of_categories))", "_____no_output_____" ], [ "# Get number of authors from text file in Google Cloud Storage\nwith file_io.FileIO(tf.gfile.Glob(filename = \"gs://{}/hybrid_recommendation/preproc/vocab_counts/author_vocab_count.txt*\".format(BUCKET))[0], mode = 'r') as ifp:\n number_of_authors = int([x for x in ifp][0])\nprint(\"number_of_authors = {}\".format(number_of_authors))", "_____no_output_____" ], [ "# Get mean months since epoch from text file in Google Cloud Storage\nwith file_io.FileIO(tf.gfile.Glob(filename = \"gs://{}/hybrid_recommendation/preproc/vocab_counts/months_since_epoch_mean.txt*\".format(BUCKET))[0], mode = 'r') as ifp:\n mean_months_since_epoch = float([x for x in ifp][0])\nprint(\"mean_months_since_epoch = {}\".format(mean_months_since_epoch))", "_____no_output_____" ], [ "# Determine CSV and label columns\nNON_FACTOR_COLUMNS = 'next_content_id,visitor_id,content_id,category,title,author,months_since_epoch'.split(',')\nFACTOR_COLUMNS = [\"user_factor_{}\".format(i) for i in range(10)] + [\"item_factor_{}\".format(i) for i in range(10)]\nCSV_COLUMNS = NON_FACTOR_COLUMNS + FACTOR_COLUMNS\nLABEL_COLUMN = 'next_content_id'\n\n# Set default values for each CSV column\nNON_FACTOR_DEFAULTS = [[\"Unknown\"],[\"Unknown\"],[\"Unknown\"],[\"Unknown\"],[\"Unknown\"],[\"Unknown\"],[mean_months_since_epoch]]\nFACTOR_DEFAULTS = [[0.0] for i in range(10)] + [[0.0] for i in range(10)] # user and item\nDEFAULTS = NON_FACTOR_DEFAULTS + FACTOR_DEFAULTS", "_____no_output_____" ] ], [ [ "Create input function for training and evaluation to read from our preprocessed CSV files.", "_____no_output_____" ] ], [ [ "# Create input function for train and eval\ndef read_dataset(filename, mode, batch_size = 512):\n def _input_fn():\n def decode_csv(value_column):\n columns = tf.decode_csv(records = value_column, record_defaults = DEFAULTS)\n features = dict(zip(CSV_COLUMNS, columns)) \n label = features.pop(LABEL_COLUMN) \n return features, label\n\n # Create list of files that match pattern\n file_list = tf.gfile.Glob(filename = filename)\n\n # Create dataset from file list\n dataset = tf.data.TextLineDataset(filenames = file_list).map(map_func = decode_csv)\n\n if mode == tf.estimator.ModeKeys.TRAIN:\n num_epochs = None # indefinitely\n dataset = dataset.shuffle(buffer_size = 10 * batch_size)\n else:\n num_epochs = 1 # end-of-input after this\n\n dataset = dataset.repeat(count = num_epochs).batch(batch_size = batch_size)\n return dataset.make_one_shot_iterator().get_next()\n return _input_fn", "_____no_output_____" ] ], [ [ "Next, we will create our feature columns using our read in features.", "_____no_output_____" ] ], [ [ "# Create feature columns to be used in model\ndef create_feature_columns(args):\n # Create content_id feature column\n content_id_column = tf.feature_column.categorical_column_with_hash_bucket(\n key = \"content_id\",\n hash_bucket_size = number_of_content_ids)\n\n # Embed content id into a lower dimensional representation\n embedded_content_column = tf.feature_column.embedding_column(\n categorical_column = content_id_column,\n dimension = args['content_id_embedding_dimensions'])\n\n # Create category feature column\n categorical_category_column = tf.feature_column.categorical_column_with_vocabulary_file(\n key = \"category\",\n vocabulary_file = tf.gfile.Glob(filename = \"gs://{}/hybrid_recommendation/preproc/vocabs/category_vocab.txt*\".format(args['bucket']))[0],\n num_oov_buckets = 1)\n\n # Convert categorical category column into indicator column so that it can be used in a DNN\n indicator_category_column = tf.feature_column.indicator_column(categorical_column = categorical_category_column)\n\n # Create title feature column using TF Hub\n embedded_title_column = hub.text_embedding_column(\n key = \"title\", \n module_spec = \"https://tfhub.dev/google/nnlm-de-dim50-with-normalization/1\",\n trainable = False)\n\n # Create author feature column\n author_column = tf.feature_column.categorical_column_with_hash_bucket(\n key = \"author\",\n hash_bucket_size = number_of_authors + 1)\n\n # Embed author into a lower dimensional representation\n embedded_author_column = tf.feature_column.embedding_column(\n categorical_column = author_column,\n dimension = args['author_embedding_dimensions'])\n\n # Create months since epoch boundaries list for our binning\n months_since_epoch_boundaries = list(range(400, 700, 20))\n\n # Create months_since_epoch feature column using raw data\n months_since_epoch_column = tf.feature_column.numeric_column(\n key = \"months_since_epoch\")\n\n # Create bucketized months_since_epoch feature column using our boundaries\n months_since_epoch_bucketized = tf.feature_column.bucketized_column(\n source_column = months_since_epoch_column,\n boundaries = months_since_epoch_boundaries)\n\n # Cross our categorical category column and bucketized months since epoch column\n crossed_months_since_category_column = tf.feature_column.crossed_column(\n keys = [categorical_category_column, months_since_epoch_bucketized],\n hash_bucket_size = len(months_since_epoch_boundaries) * (number_of_categories + 1))\n\n # Convert crossed categorical category and bucketized months since epoch column into indicator column so that it can be used in a DNN\n indicator_crossed_months_since_category_column = tf.feature_column.indicator_column(categorical_column = crossed_months_since_category_column)\n\n # Create user and item factor feature columns from our trained WALS model\n user_factors = [tf.feature_column.numeric_column(key = \"user_factor_\" + str(i)) for i in range(10)]\n item_factors = [tf.feature_column.numeric_column(key = \"item_factor_\" + str(i)) for i in range(10)]\n\n # Create list of feature columns\n feature_columns = [embedded_content_column,\n embedded_author_column,\n indicator_category_column,\n embedded_title_column,\n indicator_crossed_months_since_category_column] + user_factors + item_factors\n\n return feature_columns", "_____no_output_____" ] ], [ [ "Now we'll create our model function", "_____no_output_____" ] ], [ [ "# Create custom model function for our custom estimator\ndef model_fn(features, labels, mode, params):\n # TODO: Create neural network input layer using our feature columns defined above\n\n # TODO: Create hidden layers by looping through hidden unit list\n\n # TODO: Compute logits (1 per class) using the output of our last hidden layer\n\n # TODO: Find the predicted class indices based on the highest logit (which will result in the highest probability)\n predicted_classes = \n\n # Read in the content id vocabulary so we can tie the predicted class indices to their respective content ids\n with file_io.FileIO(tf.gfile.Glob(filename = \"gs://{}/hybrid_recommendation/preproc/vocabs/content_id_vocab.txt*\".format(BUCKET))[0], mode = 'r') as ifp:\n content_id_names = tf.constant(value = [x.rstrip() for x in ifp])\n\n # Gather predicted class names based predicted class indices\n predicted_class_names = tf.gather(params = content_id_names, indices = predicted_classes)\n\n # If the mode is prediction\n if mode == tf.estimator.ModeKeys.PREDICT:\n # Create predictions dict\n predictions_dict = {\n 'class_ids': tf.expand_dims(input = predicted_classes, axis = -1),\n 'class_names' : tf.expand_dims(input = predicted_class_names, axis = -1),\n 'probabilities': tf.nn.softmax(logits = logits),\n 'logits': logits\n }\n\n # Create export outputs\n export_outputs = {\"predict_export_outputs\": tf.estimator.export.PredictOutput(outputs = predictions_dict)}\n\n return tf.estimator.EstimatorSpec( # return early since we're done with what we need for prediction mode\n mode = mode,\n predictions = predictions_dict,\n loss = None,\n train_op = None,\n eval_metric_ops = None,\n export_outputs = export_outputs)\n\n # Continue on with training and evaluation modes\n\n # Create lookup table using our content id vocabulary\n table = tf.contrib.lookup.index_table_from_file(\n vocabulary_file = tf.gfile.Glob(filename = \"gs://{}/hybrid_recommendation/preproc/vocabs/content_id_vocab.txt*\".format(BUCKET))[0])\n\n # Look up labels from vocabulary table\n labels = table.lookup(keys = labels)\n\n # TODO: Compute loss using the correct type of softmax cross entropy since this is classification and our labels (content id indices) and probabilities are mutually exclusive\n loss = \n\n # Compute evaluation metrics of total accuracy and the accuracy of the top k classes\n accuracy = tf.metrics.accuracy(labels = labels, predictions = predicted_classes, name = 'acc_op')\n top_k_accuracy = tf.metrics.mean(values = tf.nn.in_top_k(predictions = logits, targets = labels, k = params['top_k']))\n map_at_k = tf.metrics.average_precision_at_k(labels = labels, predictions = predicted_classes, k = params['top_k'])\n\n # Put eval metrics into a dictionary\n eval_metrics = {\n 'accuracy': accuracy,\n 'top_k_accuracy': top_k_accuracy,\n 'map_at_k': map_at_k}\n\n # Create scalar summaries to see in TensorBoard\n tf.summary.scalar(name = 'accuracy', tensor = accuracy[1])\n tf.summary.scalar(name = 'top_k_accuracy', tensor = top_k_accuracy[1])\n tf.summary.scalar(name = 'map_at_k', tensor = map_at_k[1])\n\n # Create scalar summaries to see in TensorBoard\n tf.summary.scalar(name = 'accuracy', tensor = accuracy[1])\n tf.summary.scalar(name = 'top_k_accuracy', tensor = top_k_accuracy[1])\n\n # If the mode is evaluation\n if mode == tf.estimator.ModeKeys.EVAL:\n return tf.estimator.EstimatorSpec( # return early since we're done with what we need for evaluation mode\n mode = mode,\n predictions = None,\n loss = loss,\n train_op = None,\n eval_metric_ops = eval_metrics,\n export_outputs = None)\n\n # Continue on with training mode\n\n # If the mode is training\n assert mode == tf.estimator.ModeKeys.TRAIN\n\n # Create a custom optimizer\n optimizer = tf.train.AdagradOptimizer(learning_rate = params['learning_rate'])\n\n # Create train op\n train_op = optimizer.minimize(loss = loss, global_step = tf.train.get_global_step())\n\n return tf.estimator.EstimatorSpec( # final return since we're done with what we need for training mode\n mode = mode,\n predictions = None,\n loss = loss,\n train_op = train_op,\n eval_metric_ops = None,\n export_outputs = None)", "_____no_output_____" ] ], [ [ "Now create a serving input function", "_____no_output_____" ] ], [ [ "# Create serving input function\ndef serving_input_fn(): \n feature_placeholders = {\n colname : tf.placeholder(dtype = tf.string, shape = [None]) \\\n for colname in NON_FACTOR_COLUMNS[1:-1]\n }\n feature_placeholders['months_since_epoch'] = tf.placeholder(dtype = tf.float32, shape = [None])\n \n for colname in FACTOR_COLUMNS:\n feature_placeholders[colname] = tf.placeholder(dtype = tf.float32, shape = [None])\n\n features = {\n key: tf.expand_dims(tensor, -1) \\\n for key, tensor in feature_placeholders.items()\n }\n \n return tf.estimator.export.ServingInputReceiver(features, feature_placeholders)", "_____no_output_____" ] ], [ [ "Now that all of the pieces are assembled let's create and run our train and evaluate loop", "_____no_output_____" ] ], [ [ "# Create train and evaluate loop to combine all of the pieces together.\ntf.logging.set_verbosity(tf.logging.INFO)\ndef train_and_evaluate(args):\n estimator = tf.estimator.Estimator(\n model_fn = model_fn,\n model_dir = args['output_dir'],\n params={\n 'feature_columns': create_feature_columns(args),\n 'hidden_units': args['hidden_units'],\n 'n_classes': number_of_content_ids,\n 'learning_rate': args['learning_rate'],\n 'top_k': args['top_k'],\n 'bucket': args['bucket']\n })\n\n train_spec = tf.estimator.TrainSpec(\n input_fn = read_dataset(filename = args['train_data_paths'], mode = tf.estimator.ModeKeys.TRAIN, batch_size = args['batch_size']),\n max_steps = args['train_steps'])\n\n exporter = tf.estimator.LatestExporter('exporter', serving_input_fn)\n\n eval_spec = tf.estimator.EvalSpec(\n input_fn = read_dataset(filename = args['eval_data_paths'], mode = tf.estimator.ModeKeys.EVAL, batch_size = args['batch_size']),\n steps = None,\n start_delay_secs = args['start_delay_secs'],\n throttle_secs = args['throttle_secs'],\n exporters = exporter)\n\n tf.estimator.train_and_evaluate(estimator, train_spec, eval_spec)", "_____no_output_____" ] ], [ [ "Run train_and_evaluate!", "_____no_output_____" ] ], [ [ "# Call train and evaluate loop\nimport shutil\n\noutdir = 'hybrid_recommendation_trained'\nshutil.rmtree(outdir, ignore_errors = True) # start fresh each time\n\narguments = {\n 'bucket': BUCKET,\n 'train_data_paths': \"gs://{}/hybrid_recommendation/preproc/features/train.csv*\".format(BUCKET),\n 'eval_data_paths': \"gs://{}/hybrid_recommendation/preproc/features/eval.csv*\".format(BUCKET),\n 'output_dir': outdir,\n 'batch_size': 128,\n 'learning_rate': 0.1,\n 'hidden_units': [256, 128, 64],\n 'content_id_embedding_dimensions': 10,\n 'author_embedding_dimensions': 10,\n 'top_k': 10,\n 'train_steps': 1000,\n 'start_delay_secs': 30,\n 'throttle_secs': 30\n}\n\ntrain_and_evaluate(arguments)", "_____no_output_____" ] ], [ [ "## Run on module locally\n\nNow let's place our code into a python module with model.py and task.py files so that we can train using Google Cloud's ML Engine! First, let's test our module locally.", "_____no_output_____" ] ], [ [ "%writefile requirements.txt\ntensorflow_hub", "_____no_output_____" ], [ "%%bash\necho \"bucket=${BUCKET}\"\nrm -rf hybrid_recommendation_trained\nexport PYTHONPATH=${PYTHONPATH}:${PWD}/hybrid_recommendations_module\npython -m trainer.task \\\n --bucket=${BUCKET} \\\n --train_data_paths=gs://${BUCKET}/hybrid_recommendation/preproc/features/train.csv* \\\n --eval_data_paths=gs://${BUCKET}/hybrid_recommendation/preproc/features/eval.csv* \\\n --output_dir=${OUTDIR} \\\n --batch_size=128 \\\n --learning_rate=0.1 \\\n --hidden_units=\"256 128 64\" \\\n --content_id_embedding_dimensions=10 \\\n --author_embedding_dimensions=10 \\\n --top_k=10 \\\n --train_steps=1000 \\\n --start_delay_secs=30 \\\n --throttle_secs=60", "_____no_output_____" ] ], [ [ "# Run on Google Cloud ML Engine\nIf our module locally trained fine, let's now use of the power of ML Engine to scale it out on Google Cloud.", "_____no_output_____" ] ], [ [ "%%bash\nOUTDIR=gs://${BUCKET}/hybrid_recommendation/small_trained_model\nJOBNAME=hybrid_recommendation_$(date -u +%y%m%d_%H%M%S)\necho $OUTDIR $REGION $JOBNAME\ngsutil -m rm -rf $OUTDIR\ngcloud ml-engine jobs submit training $JOBNAME \\\n --region=$REGION \\\n --module-name=trainer.task \\\n --package-path=$(pwd)/hybrid_recommendations_module/trainer \\\n --job-dir=$OUTDIR \\\n --staging-bucket=gs://$BUCKET \\\n --scale-tier=STANDARD_1 \\\n --runtime-version=$TFVERSION \\\n -- \\\n --bucket=${BUCKET} \\\n --train_data_paths=gs://${BUCKET}/hybrid_recommendation/preproc/features/train.csv* \\\n --eval_data_paths=gs://${BUCKET}/hybrid_recommendation/preproc/features/eval.csv* \\\n --output_dir=${OUTDIR} \\\n --batch_size=128 \\\n --learning_rate=0.1 \\\n --hidden_units=\"256 128 64\" \\\n --content_id_embedding_dimensions=10 \\\n --author_embedding_dimensions=10 \\\n --top_k=10 \\\n --train_steps=1000 \\\n --start_delay_secs=30 \\\n --throttle_secs=30", "_____no_output_____" ] ], [ [ "Let's add some hyperparameter tuning!", "_____no_output_____" ] ], [ [ "%%writefile hyperparam.yaml\ntrainingInput:\n hyperparameters:\n goal: MAXIMIZE\n maxTrials: 5\n maxParallelTrials: 1\n hyperparameterMetricTag: accuracy\n params:\n - parameterName: batch_size\n type: INTEGER\n minValue: 8\n maxValue: 64\n scaleType: UNIT_LINEAR_SCALE\n - parameterName: learning_rate\n type: DOUBLE\n minValue: 0.01\n maxValue: 0.1\n scaleType: UNIT_LINEAR_SCALE\n - parameterName: hidden_units\n type: CATEGORICAL\n categoricalValues: ['1024 512 256', '1024 512 128', '1024 256 128', '512 256 128', '1024 512 64', '1024 256 64', '512 256 64', '1024 128 64', '512 128 64', '256 128 64', '1024 512 32', '1024 256 32', '512 256 32', '1024 128 32', '512 128 32', '256 128 32', '1024 64 32', '512 64 32', '256 64 32', '128 64 32']\n - parameterName: content_id_embedding_dimensions\n type: INTEGER\n minValue: 5\n maxValue: 250\n scaleType: UNIT_LOG_SCALE\n - parameterName: author_embedding_dimensions\n type: INTEGER\n minValue: 5\n maxValue: 30\n scaleType: UNIT_LINEAR_SCALE", "_____no_output_____" ], [ "%%bash\nOUTDIR=gs://${BUCKET}/hybrid_recommendation/hypertuning\nJOBNAME=hybrid_recommendation_$(date -u +%y%m%d_%H%M%S)\necho $OUTDIR $REGION $JOBNAME\ngsutil -m rm -rf $OUTDIR\ngcloud ml-engine jobs submit training $JOBNAME \\\n --region=$REGION \\\n --module-name=trainer.task \\\n --package-path=$(pwd)/hybrid_recommendations_module/trainer \\\n --job-dir=$OUTDIR \\\n --staging-bucket=gs://$BUCKET \\\n --scale-tier=STANDARD_1 \\\n --runtime-version=$TFVERSION \\\n --config=hyperparam.yaml \\\n -- \\\n --bucket=${BUCKET} \\\n --train_data_paths=gs://${BUCKET}/hybrid_recommendation/preproc/features/train.csv* \\\n --eval_data_paths=gs://${BUCKET}/hybrid_recommendation/preproc/features/eval.csv* \\\n --output_dir=${OUTDIR} \\\n --batch_size=128 \\\n --learning_rate=0.1 \\\n --hidden_units=\"256 128 64\" \\\n --content_id_embedding_dimensions=10 \\\n --author_embedding_dimensions=10 \\\n --top_k=10 \\\n --train_steps=1000 \\\n --start_delay_secs=30 \\\n --throttle_secs=30", "_____no_output_____" ] ], [ [ "Now that we know the best hyperparameters, run a big training job!", "_____no_output_____" ] ], [ [ "%%bash\nOUTDIR=gs://${BUCKET}/hybrid_recommendation/big_trained_model\nJOBNAME=hybrid_recommendation_$(date -u +%y%m%d_%H%M%S)\necho $OUTDIR $REGION $JOBNAME\ngsutil -m rm -rf $OUTDIR\ngcloud ml-engine jobs submit training $JOBNAME \\\n --region=$REGION \\\n --module-name=trainer.task \\\n --package-path=$(pwd)/hybrid_recommendations_module/trainer \\\n --job-dir=$OUTDIR \\\n --staging-bucket=gs://$BUCKET \\\n --scale-tier=STANDARD_1 \\\n --runtime-version=$TFVERSION \\\n -- \\\n --bucket=${BUCKET} \\\n --train_data_paths=gs://${BUCKET}/hybrid_recommendation/preproc/features/train.csv* \\\n --eval_data_paths=gs://${BUCKET}/hybrid_recommendation/preproc/features/eval.csv* \\\n --output_dir=${OUTDIR} \\\n --batch_size=128 \\\n --learning_rate=0.1 \\\n --hidden_units=\"256 128 64\" \\\n --content_id_embedding_dimensions=10 \\\n --author_embedding_dimensions=10 \\\n --top_k=10 \\\n --train_steps=10000 \\\n --start_delay_secs=30 \\\n --throttle_secs=30", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "from fastai import *\nfrom fastai.vision import *\nfrom fastai.metrics import error_rate\nimport matplotlib.pyplot as plt\nfrom fastai.utils.mem import *\n%matplotlib inline", "_____no_output_____" ] ], [ [ "###### Setting the path", "_____no_output_____" ] ], [ [ "path = Path(\"C:/Users/shahi/.fastai/data/lgg-mri-segmentation/kaggle_3m\")", "_____no_output_____" ], [ "path", "_____no_output_____" ], [ "getMask = lambda x: x.parents[0] / (x.stem + '_mask' + x.suffix)\ntempImgFile = path/\"TCGA_CS_4941_19960909/TCGA_CS_4941_19960909_1.tif\"\ntempMaskFile = getMask(tempImgFile)", "_____no_output_____" ], [ "image = open_image(tempImgFile)\nimage", "_____no_output_____" ], [ "image.shape", "_____no_output_____" ], [ "mask = open_mask(getMask(tempImgFile),div=True)\nmask", "_____no_output_____" ], [ "\nclass SegmentationLabelListWithDiv(SegmentationLabelList):\n def open(self, fn): return open_mask(fn, div=True)\nclass SegmentationItemListWithDiv(SegmentationItemList):\n _label_cls = SegmentationLabelListWithDiv", "_____no_output_____" ], [ "codes = [0,1]", "_____no_output_____" ], [ "np.random.seed(42)\nsrc = (SegmentationItemListWithDiv.from_folder(path)\n .filter_by_func(lambda x:not x.name.endswith('_mask.tif'))\n .split_by_rand_pct(0.2)\n .label_from_func(getMask, classes=codes))\n", "_____no_output_____" ], [ "data = (src.transform(get_transforms(),size=256,tfm_y=True)\n .databunch(bs=8,num_workers=0)\n .normalize(imagenet_stats))\n", "_____no_output_____" ], [ "data.train_ds", "_____no_output_____" ], [ "data.valid_ds", "_____no_output_____" ], [ "data.show_batch(2)", "_____no_output_____" ] ], [ [ "##### Building the model", "_____no_output_____" ], [ "#### Pretrained resnet 34 used for downsampling", "_____no_output_____" ] ], [ [ "learn = unet_learner(data,models.resnet34,metrics=[dice])", "_____no_output_____" ], [ "learn.lr_find()", "_____no_output_____" ], [ "learn.recorder.plot()", "_____no_output_____" ], [ "learn.fit_one_cycle(4,1e-4)", "_____no_output_____" ], [ "learn.unfreeze()\nlearn.fit_one_cycle(4,1e-4,wd=1e-2)", "_____no_output_____" ] ], [ [ "#### Resnet 34 without pretrained weights with dialation ", "_____no_output_____" ] ], [ [ "learn.save('ResnetWithPrettrained')", "_____no_output_____" ], [ "learn.summary()", "_____no_output_____" ], [ "def conv(ni,nf):\n return nn.Conv2d(ni, nf, kernel_size=3, stride=2, padding=1,dilation=2)", "_____no_output_____" ], [ "def conv2(ni,nf): return conv_layer(ni,nf)", "_____no_output_____" ], [ "models.resnet34", "_____no_output_____" ], [ "model = nn.Sequential(\n conv2(3,8),\n res_block(8),\n conv2(8,16),\n res_block(16),\n conv2(16,32),\n res_block(32),\n \n\n)", "_____no_output_____" ], [ "act_fn = nn.ReLU(inplace=True)", "_____no_output_____" ], [ "\ndef presnet(block, n_layers, name, pre=False, **kwargs):\n model = PResNet(block, n_layers, **kwargs)\n #if pre: model.load_state_dict(model_zoo.load_url(model_urls[name]))\n if pre: model.load_state_dict(torch.load(model_urls[name]))\n return model\n\n\ndef presnet18(pretrained=False, **kwargs):\n return presnet(BasicBlock, [2, 2, 2, 2], 'presnet18', pre=pretrained, **kwargs)", "_____no_output_____" ], [ "class ResBlock(nn.Module):\n def __init__(self, nf):\n super().__init__()\n self.conv1 = conv_layer(nf,nf)\n self.conv2 = conv_layer(nf,nf)\n \n def forward(self, x): return x + self.conv2(self.conv1(x))", "_____no_output_____" ], [ "model3 = nn.Sequential(\n conv2(3, 8),\n res_block(8),\n conv2(8, 16),\n res_block(16),\n conv2(16, 32),\n res_block(32),\n conv2(32, 16),\n res_block(16),\n conv2(16, 10),\n )", "_____no_output_____" ], [ "model3", "_____no_output_____" ], [ "class ResBlock(nn.Module):\n def __init__(self, nf):\n super().__init__()\n self.conv1 = conv_layer(nf,nf)\n self.conv2 = conv_layer(nf,nf)\n \n def forward(self, x): return x + self.conv2(self.conv1(x))", "_____no_output_____" ], [ "learn3 = unet_learner(data,drn_d_38,metrics=[dice],pretrained=True)", "_____no_output_____" ], [ "learn3.lr_find()", "_____no_output_____" ], [ "learn3.recorder.plot()", "_____no_output_____" ], [ "learn3.fit_one_cycle(4,1e-4,wd=1e-1)", "_____no_output_____" ], [ "learn3.unfreeze()\nlearn3.fit_one_cycle(4,1e-4,wd=1e-2)", "_____no_output_____" ], [ "learn3.save('ResnetWithPretrainedDilation')", "_____no_output_____" ], [ "learn3.summary()", "_____no_output_____" ], [ "import pdb\n\nimport torch.nn as nn\nimport math\nimport torch.utils.model_zoo as model_zoo\n\nBatchNorm = nn.BatchNorm2d\n\n\n# __all__ = ['DRN', 'drn26', 'drn42', 'drn58']\n\n\nwebroot = 'http://dl.yf.io/drn/'\n\nmodel_urls = {\n 'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',\n 'drn-c-26': webroot + 'drn_c_26-ddedf421.pth',\n 'drn-c-42': webroot + 'drn_c_42-9d336e8c.pth',\n 'drn-c-58': webroot + 'drn_c_58-0a53a92c.pth',\n 'drn-d-22': webroot + 'drn_d_22-4bd2f8ea.pth',\n 'drn-d-38': webroot + 'drn_d_38-eebb45f0.pth',\n 'drn-d-54': webroot + 'drn_d_54-0e0534ff.pth',\n 'drn-d-105': webroot + 'drn_d_105-12b40979.pth'\n}\n\n\ndef conv3x3(in_planes, out_planes, stride=1, padding=1, dilation=1):\n return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,\n padding=padding, bias=False, dilation=dilation)\n\n\nclass BasicBlock(nn.Module):\n expansion = 1\n\n def __init__(self, inplanes, planes, stride=1, downsample=None,\n dilation=(1, 1), residual=True):\n super(BasicBlock, self).__init__()\n self.conv1 = conv3x3(inplanes, planes, stride,\n padding=dilation[0], dilation=dilation[0])\n self.bn1 = BatchNorm(planes)\n self.relu = nn.ReLU(inplace=True)\n self.conv2 = conv3x3(planes, planes,\n padding=dilation[1], dilation=dilation[1])\n self.bn2 = BatchNorm(planes)\n self.downsample = downsample\n self.stride = stride\n self.residual = residual\n\n def forward(self, x):\n residual = x\n\n out = self.conv1(x)\n out = self.bn1(out)\n out = self.relu(out)\n\n out = self.conv2(out)\n out = self.bn2(out)\n\n if self.downsample is not None:\n residual = self.downsample(x)\n if self.residual:\n out += residual\n out = self.relu(out)\n\n return out\n\n\nclass Bottleneck(nn.Module):\n expansion = 4\n\n def __init__(self, inplanes, planes, stride=1, downsample=None,\n dilation=(1, 1), residual=True):\n super(Bottleneck, self).__init__()\n self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, bias=False)\n self.bn1 = BatchNorm(planes)\n self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride,\n padding=dilation[1], bias=False,\n dilation=dilation[1])\n self.bn2 = BatchNorm(planes)\n self.conv3 = nn.Conv2d(planes, planes * 4, kernel_size=1, bias=False)\n self.bn3 = BatchNorm(planes * 4)\n self.relu = nn.ReLU(inplace=True)\n self.downsample = downsample\n self.stride = stride\n\n def forward(self, x):\n residual = x\n\n out = self.conv1(x)\n out = self.bn1(out)\n out = self.relu(out)\n\n out = self.conv2(out)\n out = self.bn2(out)\n out = self.relu(out)\n\n out = self.conv3(out)\n out = self.bn3(out)\n\n if self.downsample is not None:\n residual = self.downsample(x)\n\n out += residual\n out = self.relu(out)\n\n return out\n\n\nclass DRN(nn.Module):\n\n def __init__(self, block, layers, num_classes=1000,\n channels=(16, 32, 64, 128, 256, 512, 512, 512),\n out_map=False, out_middle=False, pool_size=28, arch='D'):\n super(DRN, self).__init__()\n self.inplanes = channels[0]\n self.out_map = out_map\n self.out_dim = channels[-1]\n self.out_middle = out_middle\n self.arch = arch\n\n if arch == 'C':\n self.conv1 = nn.Conv2d(3, channels[0], kernel_size=7, stride=1,\n padding=3, bias=False)\n self.bn1 = BatchNorm(channels[0])\n self.relu = nn.ReLU(inplace=True)\n\n self.layer1 = self._make_layer(\n BasicBlock, channels[0], layers[0], stride=1)\n self.layer2 = self._make_layer(\n BasicBlock, channels[1], layers[1], stride=2)\n elif arch == 'D':\n self.layer0 = nn.Sequential(\n nn.Conv2d(3, channels[0], kernel_size=7, stride=1, padding=3,\n bias=False),\n BatchNorm(channels[0]),\n nn.ReLU(inplace=True)\n )\n\n self.layer1 = self._make_conv_layers(\n channels[0], layers[0], stride=1)\n self.layer2 = self._make_conv_layers(\n channels[1], layers[1], stride=2)\n\n self.layer3 = self._make_layer(block, channels[2], layers[2], stride=2)\n self.layer4 = self._make_layer(block, channels[3], layers[3], stride=2)\n self.layer5 = self._make_layer(block, channels[4], layers[4],\n dilation=2, new_level=False)\n self.layer6 = None if layers[5] == 0 else \\\n self._make_layer(block, channels[5], layers[5], dilation=4,\n new_level=False)\n\n if arch == 'C':\n self.layer7 = None if layers[6] == 0 else \\\n self._make_layer(BasicBlock, channels[6], layers[6], dilation=2,\n new_level=False, residual=False)\n self.layer8 = None if layers[7] == 0 else \\\n self._make_layer(BasicBlock, channels[7], layers[7], dilation=1,\n new_level=False, residual=False)\n elif arch == 'D':\n self.layer7 = None if layers[6] == 0 else \\\n self._make_conv_layers(channels[6], layers[6], dilation=2)\n self.layer8 = None if layers[7] == 0 else \\\n self._make_conv_layers(channels[7], layers[7], dilation=1)\n\n if num_classes > 0:\n self.avgpool = nn.AvgPool2d(pool_size)\n self.fc = nn.Conv2d(self.out_dim, num_classes, kernel_size=1,\n stride=1, padding=0, bias=True)\n for m in self.modules():\n if isinstance(m, nn.Conv2d):\n n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels\n m.weight.data.normal_(0, math.sqrt(2. / n))\n elif isinstance(m, BatchNorm):\n m.weight.data.fill_(1)\n m.bias.data.zero_()\n\n def _make_layer(self, block, planes, blocks, stride=1, dilation=1,\n new_level=True, residual=True):\n assert dilation == 1 or dilation % 2 == 0\n downsample = None\n if stride != 1 or self.inplanes != planes * block.expansion:\n downsample = nn.Sequential(\n nn.Conv2d(self.inplanes, planes * block.expansion,\n kernel_size=1, stride=stride, bias=False),\n BatchNorm(planes * block.expansion),\n )\n\n layers = list()\n layers.append(block(\n self.inplanes, planes, stride, downsample,\n dilation=(1, 1) if dilation == 1 else (\n dilation // 2 if new_level else dilation, dilation),\n residual=residual))\n self.inplanes = planes * block.expansion\n for i in range(1, blocks):\n layers.append(block(self.inplanes, planes, residual=residual,\n dilation=(dilation, dilation)))\n\n return nn.Sequential(*layers)\n\n def _make_conv_layers(self, channels, convs, stride=1, dilation=1):\n modules = []\n for i in range(convs):\n modules.extend([\n nn.Conv2d(self.inplanes, channels, kernel_size=3,\n stride=stride if i == 0 else 1,\n padding=dilation, bias=False, dilation=dilation),\n BatchNorm(channels),\n nn.ReLU(inplace=True)])\n self.inplanes = channels\n return nn.Sequential(*modules)\n\n def forward(self, x):\n y = list()\n\n if self.arch == 'C':\n x = self.conv1(x)\n x = self.bn1(x)\n x = self.relu(x)\n elif self.arch == 'D':\n x = self.layer0(x)\n\n x = self.layer1(x)\n y.append(x)\n x = self.layer2(x)\n y.append(x)\n\n x = self.layer3(x)\n y.append(x)\n\n x = self.layer4(x)\n y.append(x)\n\n x = self.layer5(x)\n y.append(x)\n\n if self.layer6 is not None:\n x = self.layer6(x)\n y.append(x)\n\n if self.layer7 is not None:\n x = self.layer7(x)\n y.append(x)\n\n if self.layer8 is not None:\n x = self.layer8(x)\n y.append(x)\n\n if self.out_map:\n x = self.fc(x)\n else:\n x = self.avgpool(x)\n x = self.fc(x)\n x = x.view(x.size(0), -1)\n\n if self.out_middle:\n return x, y\n else:\n return x\n\n\nclass DRN_A(nn.Module):\n\n def __init__(self, block, layers, num_classes=1000):\n self.inplanes = 64\n super(DRN_A, self).__init__()\n self.out_dim = 512 * block.expansion\n self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3,\n bias=False)\n self.bn1 = nn.BatchNorm2d(64)\n self.relu = nn.ReLU(inplace=True)\n self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)\n self.layer1 = self._make_layer(block, 64, layers[0])\n self.layer2 = self._make_layer(block, 128, layers[1], stride=2)\n self.layer3 = self._make_layer(block, 256, layers[2], stride=1,\n dilation=2)\n self.layer4 = self._make_layer(block, 512, layers[3], stride=1,\n dilation=4)\n self.avgpool = nn.AvgPool2d(28, stride=1)\n self.fc = nn.Linear(512 * block.expansion, num_classes)\n\n for m in self.modules():\n if isinstance(m, nn.Conv2d):\n n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels\n m.weight.data.normal_(0, math.sqrt(2. / n))\n elif isinstance(m, BatchNorm):\n m.weight.data.fill_(1)\n m.bias.data.zero_()\n\n # for m in self.modules():\n # if isinstance(m, nn.Conv2d):\n # nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')\n # elif isinstance(m, nn.BatchNorm2d):\n # nn.init.constant_(m.weight, 1)\n # nn.init.constant_(m.bias, 0)\n\n def _make_layer(self, block, planes, blocks, stride=1, dilation=1):\n downsample = None\n if stride != 1 or self.inplanes != planes * block.expansion:\n downsample = nn.Sequential(\n nn.Conv2d(self.inplanes, planes * block.expansion,\n kernel_size=1, stride=stride, bias=False),\n nn.BatchNorm2d(planes * block.expansion),\n )\n\n layers = []\n layers.append(block(self.inplanes, planes, stride, downsample))\n self.inplanes = planes * block.expansion\n for i in range(1, blocks):\n layers.append(block(self.inplanes, planes,\n dilation=(dilation, dilation)))\n\n return nn.Sequential(*layers)\n\n def forward(self, x):\n x = self.conv1(x)\n x = self.bn1(x)\n x = self.relu(x)\n x = self.maxpool(x)\n\n x = self.layer1(x)\n x = self.layer2(x)\n x = self.layer3(x)\n x = self.layer4(x)\n\n x = self.avgpool(x)\n x = x.view(x.size(0), -1)\n x = self.fc(x)\n\n return x\n\n\ndef drn_a_50(pretrained=False, **kwargs):\n model = DRN_A(Bottleneck, [3, 4, 6, 3], **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['resnet50']))\n return model\n\n\ndef drn_c_26(pretrained=False, **kwargs):\n model = DRN(BasicBlock, [1, 1, 2, 2, 2, 2, 1, 1], arch='C', **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['drn-c-26']))\n return model\n\n\ndef drn_c_42(pretrained=False, **kwargs):\n model = DRN(BasicBlock, [1, 1, 3, 4, 6, 3, 1, 1], arch='C', **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['drn-c-42']))\n return model\n\n\ndef drn_c_58(pretrained=False, **kwargs):\n model = DRN(Bottleneck, [1, 1, 3, 4, 6, 3, 1, 1], arch='C', **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['drn-c-58']))\n return model\n\n\ndef drn_d_22(pretrained=False, **kwargs):\n model = DRN(BasicBlock, [1, 1, 2, 2, 2, 2, 1, 1], arch='D', **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['drn-d-22']))\n return model\n\n\ndef drn_d_24(pretrained=False, **kwargs):\n model = DRN(BasicBlock, [1, 1, 2, 2, 2, 2, 2, 2], arch='D', **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['drn-d-24']))\n return model\n\n\ndef drn_d_38(pretrained=False, **kwargs):\n model = DRN(BasicBlock, [1, 1, 3, 4, 6, 3, 1, 1], arch='D', **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['drn-d-38']))\n return model\n\n\ndef drn_d_40(pretrained=False, **kwargs):\n model = DRN(BasicBlock, [1, 1, 3, 4, 6, 3, 2, 2], arch='D', **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['drn-d-40']))\n return model\n\n\ndef drn_d_54(pretrained=False, **kwargs):\n model = DRN(Bottleneck, [1, 1, 3, 4, 6, 3, 1, 1], arch='D', **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['drn-d-54']))\n return model\n\n\ndef drn_d_56(pretrained=False, **kwargs):\n model = DRN(Bottleneck, [1, 1, 3, 4, 6, 3, 2, 2], arch='D', **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['drn-d-56']))\n return model\n\n\ndef drn_d_105(pretrained=False, **kwargs):\n model = DRN(Bottleneck, [1, 1, 3, 4, 23, 3, 1, 1], arch='D', **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['drn-d-105']))\n return model\n\n\ndef drn_d_107(pretrained=False, **kwargs):\n model = DRN(Bottleneck, [1, 1, 3, 4, 23, 3, 2, 2], arch='D', **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['drn-d-107']))\n return model", "_____no_output_____" ], [ "import torch.nn as nn\nimport torch.utils.model_zoo as model_zoo\n\n\n__all__ = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101',\n 'resnet152']\n\n\nmodel_urls = {\n 'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',\n 'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',\n 'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',\n 'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',\n 'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',\n}\n\n\ndef conv3x3(in_planes, out_planes, stride=1,padding=1,dilation=1):\n \"\"\"3x3 convolution with padding\"\"\"\n return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,\n padding=1, bias=False,dilation=dilation)\n\n\ndef conv1x1(in_planes, out_planes, stride=1):\n \"\"\"1x1 convolution\"\"\"\n return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)\n\n\nclass BasicBlock(nn.Module):\n expansion = 1\n\n def __init__(self, inplanes, planes, stride=1, downsample=None,dilation=(1,1)):\n super(BasicBlock, self).__init__()\n self.conv1 = conv3x3(inplanes, planes, stride,padding=dilation[0],dilation=dilation[1])\n self.bn1 = nn.BatchNorm2d(planes)\n self.relu = nn.ReLU(inplace=True)\n self.conv2 = conv3x3(planes, planes,padding=dilation[1],dilation=dilation[1])\n self.bn2 = nn.BatchNorm2d(planes)\n self.downsample = downsample\n self.stride = stride\n \n\n def forward(self, x):\n identity = x\n\n out = self.conv1(x)\n out = self.bn1(out)\n out = self.relu(out)\n\n out = self.conv2(out)\n out = self.bn2(out)\n\n if self.downsample is not None:\n identity = self.downsample(x)\n\n out += identity\n out = self.relu(out)\n\n return out\n\n\nclass Bottleneck(nn.Module):\n expansion = 4\n\n def __init__(self, inplanes, planes, stride=1, downsample=None,dilation=(1,1)):\n super(Bottleneck, self).__init__()\n self.conv1 = conv1x1(inplanes, planes)\n self.bn1 = nn.BatchNorm2d(planes)\n self.conv2 = conv3x3(planes, planes, stride)\n self.bn2 = nn.BatchNorm2d(planes)\n self.conv3 = conv1x1(planes, planes * self.expansion)\n self.bn3 = nn.BatchNorm2d(planes * self.expansion)\n self.relu = nn.ReLU(inplace=True)\n self.downsample = downsample\n self.stride = stride\n\n def forward(self, x):\n identity = x\n\n out = self.conv1(x)\n out = self.bn1(out)\n out = self.relu(out)\n\n out = self.conv2(out)\n out = self.bn2(out)\n out = self.relu(out)\n\n out = self.conv3(out)\n out = self.bn3(out)\n\n if self.downsample is not None:\n identity = self.downsample(x)\n\n out += identity\n out = self.relu(out)\n\n return out\n\n\nclass ResNet(nn.Module):\n\n def __init__(self, block, layers, num_classes=1000, zero_init_residual=False):\n super(ResNet, self).__init__()\n self.inplanes = 64\n self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3,\n bias=False)\n self.bn1 = nn.BatchNorm2d(64)\n self.relu = nn.ReLU(inplace=True)\n self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)\n self.layer1 = self._make_layer(block, 64, layers[0])\n self.layer2 = self._make_layer(block, 128, layers[1], stride=2)\n self.layer3 = self._make_layer(block, 256, layers[2], stride=2)\n self.layer4 = self._make_layer(block, 512, layers[3], stride=2)\n self.avgpool = nn.AdaptiveAvgPool2d((1, 1))\n self.fc = nn.Linear(512 * block.expansion, num_classes)\n\n for m in self.modules():\n if isinstance(m, nn.Conv2d):\n nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')\n elif isinstance(m, nn.BatchNorm2d):\n nn.init.constant_(m.weight, 1)\n nn.init.constant_(m.bias, 0)\n\n # Zero-initialize the last BN in each residual branch,\n # so that the residual branch starts with zeros, and each residual block behaves like an identity.\n # This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677\n if zero_init_residual:\n for m in self.modules():\n if isinstance(m, Bottleneck):\n nn.init.constant_(m.bn3.weight, 0)\n elif isinstance(m, BasicBlock):\n nn.init.constant_(m.bn2.weight, 0)\n\n def _make_layer(self, block, planes, blocks, stride=1):\n downsample = None\n if stride != 1 or self.inplanes != planes * block.expansion:\n downsample = nn.Sequential(\n conv1x1(self.inplanes, planes * block.expansion, stride),\n nn.BatchNorm2d(planes * block.expansion),\n )\n\n layers = []\n layers.append(block(self.inplanes, planes, stride, downsample))\n self.inplanes = planes * block.expansion\n for _ in range(1, blocks):\n layers.append(block(self.inplanes, planes))\n\n return nn.Sequential(*layers)\n\n def forward(self, x):\n x = self.conv1(x)\n x = self.bn1(x)\n x = self.relu(x)\n x = self.maxpool(x)\n\n x = self.layer1(x)\n x = self.layer2(x)\n x = self.layer3(x)\n x = self.layer4(x)\n\n x = self.avgpool(x)\n x = x.view(x.size(0), -1)\n x = self.fc(x)\n\n return x\n\n\ndef resnet18(pretrained=False, **kwargs):\n \"\"\"Constructs a ResNet-18 model.\n\n Args:\n pretrained (bool): If True, returns a model pre-trained on ImageNet\n \"\"\"\n model = ResNet(BasicBlock, [2, 2, 2, 2], **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['resnet18']))\n return model\n\n\ndef resnet34(pretrained=False, **kwargs):\n \"\"\"Constructs a ResNet-34 model.\n\n Args:\n pretrained (bool): If True, returns a model pre-trained on ImageNet\n \"\"\"\n model = ResNet(BasicBlock, [3, 4, 6, 3], **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['resnet34']))\n return model\n\n\ndef resnet50(pretrained=False, **kwargs):\n \"\"\"Constructs a ResNet-50 model.\n\n Args:\n pretrained (bool): If True, returns a model pre-trained on ImageNet\n \"\"\"\n model = ResNet(Bottleneck, [3, 4, 6, 3], **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['resnet50']))\n return model\n\n\ndef resnet101(pretrained=False, **kwargs):\n \"\"\"Constructs a ResNet-101 model.\n\n Args:\n pretrained (bool): If True, returns a model pre-trained on ImageNet\n \"\"\"\n model = ResNet(Bottleneck, [3, 4, 23, 3], **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['resnet101']))\n return model\n\n\ndef resnet152(pretrained=False, **kwargs):\n \"\"\"Constructs a ResNet-152 model.\n\n Args:\n pretrained (bool): If True, returns a model pre-trained on ImageNet\n \"\"\"\n model = ResNet(Bottleneck, [3, 8, 36, 3], **kwargs)\n if pretrained:\n model.load_state_dict(model_zoo.load_url(model_urls['resnet152']))\n return model\n", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Supervised Learning: Finding Donors for *CharityML*\n\n> Udacity Machine Learning Engineer Nanodegree: _Project 2_\n>\n> Author: _Ke Zhang_\n>\n> Submission Date: _2017-04-30_ (Revision 3)", "_____no_output_____" ], [ "## Content\n\n- [Getting Started](#Getting-Started)\n- [Exploring the Data](#Exploring-the-Data)\n- [Preparing the Data](#Preparing-the-Data)\n- [Evaluating Model Performance](#Evaluating-Model-Performance)\n- [Improving Results](#Improving-Results)\n- [Feature Importance](#Feature-Importance)\n- [References](#References)\n- [Reproduction Environment](#Reproduction-Environment)", "_____no_output_____" ], [ "## Getting Started\n\nIn this project, you will employ several supervised algorithms of your choice to accurately model individuals' income using data collected from the 1994 U.S. Census. You will then choose the best candidate algorithm from preliminary results and further optimize this algorithm to best model the data. Your goal with this implementation is to construct a model that accurately predicts whether an individual makes more than $50,000. This sort of task can arise in a non-profit setting, where organizations survive on donations. Understanding an individual's income can help a non-profit better understand how large of a donation to request, or whether or not they should reach out to begin with. While it can be difficult to determine an individual's general income bracket directly from public sources, we can (as we will see) infer this value from other publically available features. \n\nThe dataset for this project originates from the [UCI Machine Learning Repository](https://archive.ics.uci.edu/ml/datasets/Census+Income). The datset was donated by Ron Kohavi and Barry Becker, after being published in the article _\"Scaling Up the Accuracy of Naive-Bayes Classifiers: A Decision-Tree Hybrid\"_. You can find the article by Ron Kohavi [online](https://www.aaai.org/Papers/KDD/1996/KDD96-033.pdf). The data we investigate here consists of small changes to the original dataset, such as removing the `'fnlwgt'` feature and records with missing or ill-formatted entries.", "_____no_output_____" ], [ "----\n## Exploring the Data\nRun the code cell below to load necessary Python libraries and load the census data. Note that the last column from this dataset, `'income'`, will be our target label (whether an individual makes more than, or at most, $50,000 annually). All other columns are features about each individual in the census database.", "_____no_output_____" ] ], [ [ "# Import libraries necessary for this project\nimport numpy as np\nimport pandas as pd\nfrom time import time\nfrom IPython.display import display # Allows the use of display() for DataFrames\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n\n# Import supplementary visualization code visuals.py\nimport visuals as vs\n\n#sklearn makes lots of deprecation warnings...\nimport warnings\nwarnings.filterwarnings(\"ignore\", category=DeprecationWarning)\n\n# Pretty display for notebooks\n%matplotlib inline\nsns.set(style='white', palette='muted', color_codes=True)\nsns.set_context('notebook', font_scale=1.2, rc={'lines.linewidth': 1.2})\n\n# Load the Census dataset\ndata = pd.read_csv(\"census.csv\")\n\n# Success - Display the first record\ndisplay(data.head(n=1))", "_____no_output_____" ] ], [ [ "### Implementation: Data Exploration\nA cursory investigation of the dataset will determine how many individuals fit into either group, and will tell us about the percentage of these individuals making more than \\$50,000. In the code cell below, you will need to compute the following:\n- The total number of records, `'n_records'`\n- The number of individuals making more than \\$50,000 annually, `'n_greater_50k'`.\n- The number of individuals making at most \\$50,000 annually, `'n_at_most_50k'`.\n- The percentage of individuals making more than \\$50,000 annually, `'greater_percent'`.\n\n**Hint:** You may need to look at the table above to understand how the `'income'` entries are formatted. ", "_____no_output_____" ] ], [ [ "# Total number of records\nn_records = data.shape[0]\n\n# Number of records where individual's income is more than $50,000\nn_greater_50k = data[data['income'] == '>50K'].shape[0]\n\n# Number of records where individual's income is at most $50,000\nn_at_most_50k = data[data['income'] == '<=50K'].shape[0]\n\n# Percentage of individuals whose income is more than $50,000\ngreater_percent = n_greater_50k / (n_records / 100.0)\n\n# Print the results\nprint \"Total number of records: {}\".format(n_records)\nprint \"Individuals making more than $50,000: {}\".format(n_greater_50k)\nprint \"Individuals making at most $50,000: {}\".format(n_at_most_50k)\nprint \"Percentage of individuals making more than $50,000: {:.2f}%\".format(greater_percent)", "Total number of records: 45222\nIndividuals making more than $50,000: 11208\nIndividuals making at most $50,000: 34014\nPercentage of individuals making more than $50,000: 24.78%\n" ] ], [ [ "----\n## Preparing the Data\nBefore data can be used as input for machine learning algorithms, it often must be cleaned, formatted, and restructured — this is typically known as **preprocessing**. Fortunately, for this dataset, there are no invalid or missing entries we must deal with, however, there are some qualities about certain features that must be adjusted. This preprocessing can help tremendously with the outcome and predictive power of nearly all learning algorithms.", "_____no_output_____" ], [ "### Transforming Skewed Continuous Features\nA dataset may sometimes contain at least one feature whose values tend to lie near a single number, but will also have a non-trivial number of vastly larger or smaller values than that single number. Algorithms can be sensitive to such distributions of values and can underperform if the range is not properly normalized. With the census dataset two features fit this description: '`capital-gain'` and `'capital-loss'`. \n\nRun the code cell below to plot a histogram of these two features. Note the range of the values present and how they are distributed.", "_____no_output_____" ] ], [ [ "# Split the data into features and target label\nincome_raw = data['income']\nfeatures_raw = data.drop('income', axis = 1)\n\n# Visualize skewed continuous features of original data\nvs.distribution(data)", "_____no_output_____" ] ], [ [ "For highly-skewed feature distributions such as `'capital-gain'` and `'capital-loss'`, it is common practice to apply a <a href=\"https://en.wikipedia.org/wiki/Data_transformation_(statistics)\">logarithmic transformation</a> on the data so that the very large and very small values do not negatively affect the performance of a learning algorithm. Using a logarithmic transformation significantly reduces the range of values caused by outliers. Care must be taken when applying this transformation however: The logarithm of `0` is undefined, so we must translate the values by a small amount above `0` to apply the the logarithm successfully.\n\nRun the code cell below to perform a transformation on the data and visualize the results. Again, note the range of values and how they are distributed. ", "_____no_output_____" ] ], [ [ "# Log-transform the skewed features\nskewed = ['capital-gain', 'capital-loss']\nfeatures_raw[skewed] = data[skewed].apply(lambda x: np.log(x + 1))\n\n# Visualize the new log distributions\nvs.distribution(features_raw, transformed = True)", "_____no_output_____" ] ], [ [ "### Normalizing Numerical Features\nIn addition to performing transformations on features that are highly skewed, it is often good practice to perform some type of scaling on numerical features. Applying a scaling to the data does not change the shape of each feature's distribution (such as `'capital-gain'` or `'capital-loss'` above); however, normalization ensures that each feature is treated equally when applying supervised learners. Note that once scaling is applied, observing the data in its raw form will no longer have the same original meaning, as exampled below.\n\nRun the code cell below to normalize each numerical feature. We will use [`sklearn.preprocessing.MinMaxScaler`](http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.MinMaxScaler.html) for this.", "_____no_output_____" ] ], [ [ "# Import sklearn.preprocessing.StandardScaler\nfrom sklearn.preprocessing import MinMaxScaler\n\n# Initialize a scaler, then apply it to the features\nscaler = MinMaxScaler()\nnumerical = ['age', 'education-num', 'capital-gain', 'capital-loss', 'hours-per-week']\nfeatures_raw[numerical] = scaler.fit_transform(data[numerical])\n\n# Show an example of a record with scaling applied\ndisplay(features_raw.head(n = 1))", "_____no_output_____" ] ], [ [ "### Implementation: Data Preprocessing\n\nFrom the table in **Exploring the Data** above, we can see there are several features for each record that are non-numeric. Typically, learning algorithms expect input to be numeric, which requires that non-numeric features (called *categorical variables*) be converted. One popular way to convert categorical variables is by using the **one-hot encoding** scheme. One-hot encoding creates a _\"dummy\"_ variable for each possible category of each non-numeric feature. For example, assume `someFeature` has three possible entries: `A`, `B`, or `C`. We then encode this feature into `someFeature_A`, `someFeature_B` and `someFeature_C`.\n\n| | someFeature | | someFeature_A | someFeature_B | someFeature_C |\n| :-: | :-: | | :-: | :-: | :-: |\n| 0 | B | | 0 | 1 | 0 |\n| 1 | C | ----> one-hot encode ----> | 0 | 0 | 1 |\n| 2 | A | | 1 | 0 | 0 |\n\nAdditionally, as with the non-numeric features, we need to convert the non-numeric target label, `'income'` to numerical values for the learning algorithm to work. Since there are only two possible categories for this label (\"<=50K\" and \">50K\"), we can avoid using one-hot encoding and simply encode these two categories as `0` and `1`, respectively. In code cell below, you will need to implement the following:\n - Use [`pandas.get_dummies()`](http://pandas.pydata.org/pandas-docs/stable/generated/pandas.get_dummies.html?highlight=get_dummies#pandas.get_dummies) to perform one-hot encoding on the `'features_raw'` data.\n - Convert the target label `'income_raw'` to numerical entries.\n - Set records with \"<=50K\" to `0` and records with \">50K\" to `1`.", "_____no_output_____" ] ], [ [ "# One-hot encode the 'features_raw' data using pandas.get_dummies()\nfeatures = pd.get_dummies(features_raw)\n\n# Encode the 'income_raw' data to numerical values\nincome = income_raw.apply(lambda x: 1 if x == '>50K' else 0)\n\n# Print the number of features after one-hot encoding\nencoded = list(features.columns)\nprint \"{} total features after one-hot encoding.\".format(len(encoded))\n\n# Uncomment the following line to see the encoded feature names\nprint encoded", "103 total features after one-hot encoding.\n['age', 'education-num', 'capital-gain', 'capital-loss', 'hours-per-week', 'workclass_ Federal-gov', 'workclass_ Local-gov', 'workclass_ Private', 'workclass_ Self-emp-inc', 'workclass_ Self-emp-not-inc', 'workclass_ State-gov', 'workclass_ Without-pay', 'education_level_ 10th', 'education_level_ 11th', 'education_level_ 12th', 'education_level_ 1st-4th', 'education_level_ 5th-6th', 'education_level_ 7th-8th', 'education_level_ 9th', 'education_level_ Assoc-acdm', 'education_level_ Assoc-voc', 'education_level_ Bachelors', 'education_level_ Doctorate', 'education_level_ HS-grad', 'education_level_ Masters', 'education_level_ Preschool', 'education_level_ Prof-school', 'education_level_ Some-college', 'marital-status_ Divorced', 'marital-status_ Married-AF-spouse', 'marital-status_ Married-civ-spouse', 'marital-status_ Married-spouse-absent', 'marital-status_ Never-married', 'marital-status_ Separated', 'marital-status_ Widowed', 'occupation_ Adm-clerical', 'occupation_ Armed-Forces', 'occupation_ Craft-repair', 'occupation_ Exec-managerial', 'occupation_ Farming-fishing', 'occupation_ Handlers-cleaners', 'occupation_ Machine-op-inspct', 'occupation_ Other-service', 'occupation_ Priv-house-serv', 'occupation_ Prof-specialty', 'occupation_ Protective-serv', 'occupation_ Sales', 'occupation_ Tech-support', 'occupation_ Transport-moving', 'relationship_ Husband', 'relationship_ Not-in-family', 'relationship_ Other-relative', 'relationship_ Own-child', 'relationship_ Unmarried', 'relationship_ Wife', 'race_ Amer-Indian-Eskimo', 'race_ Asian-Pac-Islander', 'race_ Black', 'race_ Other', 'race_ White', 'sex_ Female', 'sex_ Male', 'native-country_ Cambodia', 'native-country_ Canada', 'native-country_ China', 'native-country_ Columbia', 'native-country_ Cuba', 'native-country_ Dominican-Republic', 'native-country_ Ecuador', 'native-country_ El-Salvador', 'native-country_ England', 'native-country_ France', 'native-country_ Germany', 'native-country_ Greece', 'native-country_ Guatemala', 'native-country_ Haiti', 'native-country_ Holand-Netherlands', 'native-country_ Honduras', 'native-country_ Hong', 'native-country_ Hungary', 'native-country_ India', 'native-country_ Iran', 'native-country_ Ireland', 'native-country_ Italy', 'native-country_ Jamaica', 'native-country_ Japan', 'native-country_ Laos', 'native-country_ Mexico', 'native-country_ Nicaragua', 'native-country_ Outlying-US(Guam-USVI-etc)', 'native-country_ Peru', 'native-country_ Philippines', 'native-country_ Poland', 'native-country_ Portugal', 'native-country_ Puerto-Rico', 'native-country_ Scotland', 'native-country_ South', 'native-country_ Taiwan', 'native-country_ Thailand', 'native-country_ Trinadad&Tobago', 'native-country_ United-States', 'native-country_ Vietnam', 'native-country_ Yugoslavia']\n" ] ], [ [ "### Shuffle and Split Data\nNow all _categorical variables_ have been converted into numerical features, and all numerical features have been normalized. As always, we will now split the data (both features and their labels) into training and test sets. 80% of the data will be used for training and 20% for testing.\n\nRun the code cell below to perform this split.", "_____no_output_____" ] ], [ [ "# Import train_test_split\nfrom sklearn.cross_validation import train_test_split\n\n# Split the 'features' and 'income' data into training and testing sets\nX_train, X_test, y_train, y_test = train_test_split(features, income, test_size = 0.2, random_state = 0)\n\n# Show the results of the split\nprint \"Training set has {} samples.\".format(X_train.shape[0])\nprint \"Testing set has {} samples.\".format(X_test.shape[0])", "Training set has 36177 samples.\nTesting set has 9045 samples.\n" ] ], [ [ "----\n## Evaluating Model Performance\nIn this section, we will investigate four different algorithms, and determine which is best at modeling the data. Three of these algorithms will be supervised learners of your choice, and the fourth algorithm is known as a *naive predictor*.", "_____no_output_____" ], [ "### Metrics and the Naive Predictor\n*CharityML*, equipped with their research, knows individuals that make more than \\$50,000 are most likely to donate to their charity. Because of this, *CharityML* is particularly interested in predicting who makes more than \\$50,000 accurately. It would seem that using **accuracy** as a metric for evaluating a particular model's performace would be appropriate. Additionally, identifying someone that *does not* make more than \\$50,000 as someone who does would be detrimental to *CharityML*, since they are looking to find individuals willing to donate. Therefore, a model's ability to precisely predict those that make more than \\$50,000 is *more important* than the model's ability to **recall** those individuals. We can use **F-beta score** as a metric that considers both precision and recall:\n\n$$ F_{\\beta} = (1 + \\beta^2) \\cdot \\frac{precision \\cdot recall}{\\left( \\beta^2 \\cdot precision \\right) + recall} $$\n\nIn particular, when $\\beta = 0.5$, more emphasis is placed on precision. This is called the **F$_{0.5}$ score** (or F-score for simplicity).\n\nLooking at the distribution of classes (those who make at most \\$50,000, and those who make more), it's clear most individuals do not make more than \\$50,000. This can greatly affect **accuracy**, since we could simply say *\"this person does not make more than \\$50,000\"* and generally be right, without ever looking at the data! Making such a statement would be called **naive**, since we have not considered any information to substantiate the claim. It is always important to consider the *naive prediction* for your data, to help establish a benchmark for whether a model is performing well. That been said, using that prediction would be pointless: If we predicted all people made less than \\$50,000, *CharityML* would identify no one as donors. ", "_____no_output_____" ], [ "### Question 1 - Naive Predictor Performace\n*If we chose a model that always predicted an individual made more than \\$50,000, what would that model's accuracy and F-score be on this dataset?* \n**Note:** You must use the code cell below and assign your results to `'accuracy'` and `'fscore'` to be used later.", "_____no_output_____" ] ], [ [ "# Calculate accuracy\naccuracy = 1.0 * n_greater_50k / n_records\n\n# Calculate F-score using the formula above for beta = 0.5\nrecall = 1.0\nfscore = (\n (1 + 0.5**2) * accuracy * recall\n) / (\n 0.5**2 * accuracy + recall\n)\n\n# Print the results \nprint \"Naive Predictor: [Accuracy score: {:.4f}, F-score: {:.4f}]\".format(accuracy, fscore)", "Naive Predictor: [Accuracy score: 0.2478, F-score: 0.2917]\n" ] ], [ [ "### Supervised Learning Models\n**The following supervised learning models are currently available in** [`scikit-learn`](http://scikit-learn.org/stable/supervised_learning.html) **that you may choose from:**\n- Gaussian Naive Bayes (GaussianNB)\n- Decision Trees\n- Ensemble Methods (Bagging, AdaBoost, Random Forest, Gradient Boosting)\n- K-Nearest Neighbors (KNeighbors)\n- Stochastic Gradient Descent Classifier (SGDC)\n- Support Vector Machines (SVM)\n- Logistic Regression", "_____no_output_____" ], [ "### Question 2 - Model Application\nList three of the supervised learning models above that are appropriate for this problem that you will test on the census data. For each model chosen\n- *Describe one real-world application in industry where the model can be applied.* (You may need to do research for this — give references!)\n- *What are the strengths of the model; when does it perform well?*\n- *What are the weaknesses of the model; when does it perform poorly?*\n- *What makes this model a good candidate for the problem, given what you know about the data?*", "_____no_output_____" ], [ "**Answer: **\nTotal number of records: 45222\n\nThe algorithms we're searching here is for a supervised classification problem predicting a category using labeled data with less than 100K samples.\n* **Support Vector Machine** (SVM):\n * Real-world application: classifying proteins, protein-protein interaction\n * References: [Bioinformatics - Semi-Supervised Multi-Task Learning for Predicting Interactions between HIV-1 and Human Proteins](https://static.googleusercontent.com/media/research.google.com/de//pubs/archive/35765.pdf)\n * Strengths of the model:\n * effective in high dimensional spaces and with nonlinear relationships\n * robust to noise (because margins maximized and theoretical bounds on overfitting)\n * Weaknesses of the model:\n * requires to select a good kernel function and a number of hyperparameters such as the regularization parameter and the number of iterations\n * sensitive to feature scaling\n * model parameters are difficult to interpret\n * requires significant memory and processing power\n * tuning the regularization parameters required to avoid overfitting\n * Reasoning: *Linear SVC* is the optimal estimator following the [Scikit-Learn - Choosing the right estimator](http://scikit-learn.org/stable/tutorial/machine_learning_map/) when using less than 100K samples to solve to classification problem.\n* **Logistic Regression**:\n * Real-world application: media and advertising campaigns optimization and decision making\n * References: [Evaluating Online Ad Campaigns in a Pipeline: Causal Models At Scale](https://static.googleusercontent.com/media/research.google.com/de//pubs/archive/36552.pdf)\n * Strengths of the model:\n * simple, no user-defined parameters to experiment with unless you regularize, β is intuitive\n * fast to train and to predict\n * easy to interpret: output can be interpreted as a probability\n * pretty robust to noise with low variance and less prone to over-fitting\n * lots of ways to regularize the model\n * Weaknesses of the model:\n * unstable when one predictor could almost explain the response variable\n * often less accurate than the newer methods\n * Interpreting θ isn't straightforward\n * Reasoning: It's similar to *linear SVC*, is widely used and can be easyly implemented.\n* **K-Nearest Neighbors (KNeighbors)**:\n * Real-world application: image and video content classification\n * References: [Clustering billions of images with large scale nearest neighbor search](https://static.googleusercontent.com/media/research.google.com/de//pubs/archive/32616.pdf)\n * Strengths of the model: \n * simple and powerful\n * easy to explain\n * no training involved (\"lazy\")\n * naturally handles multiclass classification and regression\n * learns nonlinear boundaries\n * Weaknesses of the model:\n * expensive and slow to predict new instances (\"lazy\")\n * must define a meaningful distance function (preference bias)\n * need to decide on a good distance metric\n * performs poorly on high-dimensionality datasets (curse of high dimensionality) \n * Reasoning: *KNeighbors* classifier is the next option suggested on the machine learning map when *linear SVC* does work poor. Since our dataset is low-dimensional and it should generate reasonable results.\n ", "_____no_output_____" ], [ "### Implementation - Creating a Training and Predicting Pipeline\nTo properly evaluate the performance of each model you've chosen, it's important that you create a training and predicting pipeline that allows you to quickly and effectively train models using various sizes of training data and perform predictions on the testing data. Your implementation here will be used in the following section.\nIn the code block below, you will need to implement the following:\n - Import `fbeta_score` and `accuracy_score` from [`sklearn.metrics`](http://scikit-learn.org/stable/modules/classes.html#sklearn-metrics-metrics).\n - Fit the learner to the sampled training data and record the training time.\n - Perform predictions on the test data `X_test`, and also on the first 300 training points `X_train[:300]`.\n - Record the total prediction time.\n - Calculate the accuracy score for both the training subset and testing set.\n - Calculate the F-score for both the training subset and testing set.\n - Make sure that you set the `beta` parameter!", "_____no_output_____" ] ], [ [ "# Import two metrics from sklearn - fbeta_score and accuracy_score\nfrom sklearn.metrics import fbeta_score, accuracy_score\n\ndef train_predict(learner, sample_size, X_train, y_train, X_test, y_test): \n '''\n inputs:\n - learner: the learning algorithm to be trained and predicted on\n - sample_size: the size of samples (number) to be drawn from training set\n - X_train: features training set\n - y_train: income training set\n - X_test: features testing set\n - y_test: income testing set\n '''\n \n results = {}\n \n # Fit the learner to the training data using slicing with 'sample_size'\n start = time() # Get start time\n learner = learner.fit(X_train[:sample_size], y_train[:sample_size])\n end = time() # Get end time\n \n # Calculate the training time\n results['train_time'] = end - start\n \n # Get the predictions on the test set,\n # then get predictions on the first 300 training samples\n start = time() # Get start time\n predictions_test = learner.predict(X_test)\n predictions_train = learner.predict(X_train[:300])\n end = time() # Get end time\n \n # Calculate the total prediction time\n results['pred_time'] = end - start\n \n # Compute accuracy on the first 300 training samples\n results['acc_train'] = accuracy_score(y_train[:300], predictions_train)\n \n # Compute accuracy on test set\n results['acc_test'] = accuracy_score(y_test, predictions_test)\n \n # Compute F-score on the the first 300 training samples\n results['f_train'] = fbeta_score(y_train[:300], predictions_train, beta=.5)\n \n # Compute F-score on the test set\n results['f_test'] = fbeta_score(y_test, predictions_test, beta=.5)\n \n # Success\n print \"{} trained on {} samples.\".format(learner.__class__.__name__, sample_size)\n \n # Return the results\n return results", "_____no_output_____" ] ], [ [ "### Implementation: Initial Model Evaluation\nIn the code cell, you will need to implement the following:\n- Import the three supervised learning models you've discussed in the previous section.\n- Initialize the three models and store them in `'clf_A'`, `'clf_B'`, and `'clf_C'`.\n - Use a `'random_state'` for each model you use, if provided.\n - **Note:** Use the default settings for each model — you will tune one specific model in a later section.\n- Calculate the number of records equal to 1%, 10%, and 100% of the training data.\n - Store those values in `'samples_1'`, `'samples_10'`, and `'samples_100'` respectively.\n\n**Note:** Depending on which algorithms you chose, the following implementation may take some time to run!", "_____no_output_____" ] ], [ [ "# Import the three supervised learning models from sklearn\nfrom sklearn.svm import LinearSVC\nfrom sklearn.linear_model import LogisticRegression\nfrom sklearn.neighbors import KNeighborsClassifier\n\n# Initialize the three models\nclf_A = LinearSVC(random_state=42)\nclf_B = LogisticRegression(random_state=42)\nclf_C = KNeighborsClassifier()\n\n# Calculate the number of samples for 1%, 10%, and 100% of the training data\nn = len(y_train)\nsamples_1 = int(round(n / 100.0))\nsamples_10 = int(round(n / 10.0))\nsamples_100 = n\n\n# Collect results on the learners\nresults = {}\nfor clf in [clf_A, clf_B, clf_C]:\n clf_name = clf.__class__.__name__\n results[clf_name] = {}\n for i, samples in enumerate([samples_1, samples_10, samples_100]):\n results[clf_name][i] = \\\n train_predict(clf, samples, X_train, y_train, X_test, y_test)\n\n# Run metrics visualization for the three supervised learning models chosen\nvs.evaluate(results, accuracy, fscore)", "LinearSVC trained on 362 samples.\nLinearSVC trained on 3618 samples.\nLinearSVC trained on 36177 samples.\nLogisticRegression trained on 362 samples.\nLogisticRegression trained on 3618 samples.\nLogisticRegression trained on 36177 samples.\nKNeighborsClassifier trained on 362 samples.\nKNeighborsClassifier trained on 3618 samples.\nKNeighborsClassifier trained on 36177 samples.\n" ] ], [ [ "----\n## Improving Results\nIn this final section, you will choose from the three supervised learning models the *best* model to use on the student data. You will then perform a grid search optimization for the model over the entire training set (`X_train` and `y_train`) by tuning at least one parameter to improve upon the untuned model's F-score. ", "_____no_output_____" ], [ "### Question 3 - Choosing the Best Model\n*Based on the evaluation you performed earlier, in one to two paragraphs, explain to *CharityML* which of the three models you believe to be most appropriate for the task of identifying individuals that make more than \\$50,000.* \n**Hint:** Your answer should include discussion of the metrics, prediction/training time, and the algorithm's suitability for the data.", "_____no_output_____" ], [ "**Answer: **\n\n| | training time | predicting time | training set scores | testing set scores |\n|----------------------|----------------------------|-----------------|----------------|\n| **LinearSVC** | ++ | +++ | +++ | **+++** |\n| LogisticRegression | +++ | +++ | ++ | ++ |\n| KNeighborsClassifier | + | --- | +++ | + |\n\nBased on the evaluation the *linear SVC* model is the most appropriate for the task. Compared to the other two models in testing, it's both fast and has the highest scores. While *linear SVC* and *logistic regression* have almost the same accuracy, its *f-score* is slightly higher indicating that *linear SVC* has more correct positive predictions on actual positive observations.\n\nAlthough *logistic regression* has a shorter training time, it has worse *f-score* when applied to the testing set. In a real-world setting the testing scores are the metrics that do matter. The moderate longer training time can be ignored. *K-neighbors* outperforms the other two only when predicting the training scores. But with an even poorer testing score it actually shows that the model has an overfitting problem.", "_____no_output_____" ], [ "### Question 4 - Describing the Model in Layman's Terms\n*In one to two paragraphs, explain to *CharityML*, in layman's terms, how the final model chosen is supposed to work. Be sure that you are describing the major qualities of the model, such as how the model is trained and how the model makes a prediction. Avoid using advanced mathematical or technical jargon, such as describing equations or discussing the algorithm implementation.*", "_____no_output_____" ], [ "**Answer: ** \n\nAs the final model we used a classifier called *linear SVC* which assumes that the underlying data are linearly separable. In a simplified 2-dimensional space this technique attemps to find the best line that separates the two classes of points with the largest margin. In higher dimensional space, the algorithm searches for the best hyperplane following the same principle by maximizing the margin (e.g. a plane in 3-dimensional space).\n\nIn our case, in the training phase the algorithm calculates the best model to separate the different classes in the training data by a maximum margin. And with the trained model the algorithm is able to predict the unseen examples.\n", "_____no_output_____" ], [ "### Implementation: Model Tuning\nFine tune the chosen model. Use grid search (`GridSearchCV`) with at least one important parameter tuned with at least 3 different values. You will need to use the entire training set for this. In the code cell below, you will need to implement the following:\n- Import [`sklearn.grid_search.GridSearchCV`](http://scikit-learn.org/0.17/modules/generated/sklearn.grid_search.GridSearchCV.html) and [`sklearn.metrics.make_scorer`](http://scikit-learn.org/stable/modules/generated/sklearn.metrics.make_scorer.html).\n- Initialize the classifier you've chosen and store it in `clf`.\n - Set a `random_state` if one is available to the same state you set before.\n- Create a dictionary of parameters you wish to tune for the chosen model.\n - Example: `parameters = {'parameter' : [list of values]}`.\n - **Note:** Avoid tuning the `max_features` parameter of your learner if that parameter is available!\n- Use `make_scorer` to create an `fbeta_score` scoring object (with $\\beta = 0.5$).\n- Perform grid search on the classifier `clf` using the `'scorer'`, and store it in `grid_obj`.\n- Fit the grid search object to the training data (`X_train`, `y_train`), and store it in `grid_fit`.\n\n**Note:** Depending on the algorithm chosen and the parameter list, the following implementation may take some time to run!", "_____no_output_____" ] ], [ [ "# Import 'GridSearchCV', 'make_scorer', and any other necessary libraries\nfrom sklearn.grid_search import GridSearchCV\nfrom sklearn.metrics import make_scorer\n\n# Initialize the classifier\nclf = LinearSVC(random_state=42)\n\n# Create the parameters list you wish to tune\nparameters = {\n 'C': [.1, .5, 1.0, 5.0, 10.0],\n 'loss': ['hinge', 'squared_hinge'],\n 'tol': [1e-3, 1e-4, 1e-5],\n 'random_state': [0, 42, 10000]\n}\n\n# Make an fbeta_score scoring object\nscorer = make_scorer(fbeta_score, beta=.5)\n\n# Perform grid search on the classifier using 'scorer' as the scoring method\ngrid_obj = GridSearchCV(clf, parameters, scoring=scorer)\n\n# Fit the grid search object to the training data and find the optimal parameters\ngrid_fit = grid_obj.fit(X_train, y_train)\n\n# Get the estimator\nbest_clf = grid_fit.best_estimator_\n\n# Make predictions using the unoptimized and model\npredictions = (clf.fit(X_train, y_train)).predict(X_test)\nbest_predictions = best_clf.predict(X_test)\n\n# Report the before-and-afterscores\nprint \"Unoptimized model\\n------\"\nprint \"Accuracy score on testing data: {:.4f}\".format(accuracy_score(y_test, predictions))\nprint \"F-score on testing data: {:.4f}\".format(fbeta_score(y_test, predictions, beta = 0.5))\nprint \"\\nOptimized Model\\n------\"\nprint \"Final accuracy score on the testing data: {:.4f}\".format(accuracy_score(y_test, best_predictions))\nprint \"Final F-score on the testing data: {:.4f}\".format(fbeta_score(y_test, best_predictions, beta = 0.5))", "Unoptimized model\n------\nAccuracy score on testing data: 0.8507\nF-score on testing data: 0.7054\n\nOptimized Model\n------\nFinal accuracy score on the testing data: 0.8514\nFinal F-score on the testing data: 0.7063\n" ], [ "# print optimized parameters\nprint(\"Optimized params for Linear SVM: {}\".format(\n grid_fit.best_params_\n))", "Optimized params for Linear SVM: {'loss': 'squared_hinge', 'C': 10.0, 'random_state': 0, 'tol': 0.001}\n" ] ], [ [ "### Question 5 - Final Model Evaluation\n_What is your optimized model's accuracy and F-score on the testing data? Are these scores better or worse than the unoptimized model? How do the results from your optimized model compare to the naive predictor benchmarks you found earlier in **Question 1**?_ \n**Note:** Fill in the table below with your results, and then provide discussion in the **Answer** box.", "_____no_output_____" ], [ "#### Results:\n\n| Metric | Benchmark Predictor | Unoptimized Model | Optimized Model |\n| :------------: | :-----------------: | :---------------: | :-------------: | \n| Accuracy Score | .2478 | .8507 | .8514 |\n| F-score | .2917 | .7054 | .7063 |\n", "_____no_output_____" ], [ "**Answer: **\n\nThe scores of the *optimized model* are a bit better than the *unoptimized model* showing that the defaults were actually very good in the first place. Compared to the *benchmark predictor*, the optimized model is by manifold better with larger accuracy and f-scores.", "_____no_output_____" ], [ "----\n## Feature Importance\n\nAn important task when performing supervised learning on a dataset like the census data we study here is determining which features provide the most predictive power. By focusing on the relationship between only a few crucial features and the target label we simplify our understanding of the phenomenon, which is most always a useful thing to do. In the case of this project, that means we wish to identify a small number of features that most strongly predict whether an individual makes at most or more than \\$50,000.\n\nChoose a scikit-learn classifier (e.g., adaboost, random forests) that has a `feature_importance_` attribute, which is a function that ranks the importance of features according to the chosen classifier. In the next python cell fit this classifier to training set and use this attribute to determine the top 5 most important features for the census dataset.", "_____no_output_____" ], [ "### Question 6 - Feature Relevance Observation\nWhen **Exploring the Data**, it was shown there are thirteen available features for each individual on record in the census data. \n_Of these thirteen records, which five features do you believe to be most important for prediction, and in what order would you rank them and why?_", "_____no_output_____" ], [ "**Answer:**\nOf the thirteen available features we believe that the following five features are the most important for prediction and ordered them by their importance with the most important at top. From looking at the data and following the rules of our society:\n* *occupation*: different occupations have usually different salary ranges\n* *capital-gain*: the rich get richer. Capital gain is an indicator of the personal wealth.\n* *education-level*: when employed, people with higher education gets better paid\n* *hours-per-week*: part time jobs are often less paid\n* *age*: older people tends to earn more", "_____no_output_____" ], [ "### Implementation - Extracting Feature Importance\nChoose a `scikit-learn` supervised learning algorithm that has a `feature_importance_` attribute availble for it. This attribute is a function that ranks the importance of each feature when making predictions based on the chosen algorithm.\n\nIn the code cell below, you will need to implement the following:\n - Import a supervised learning model from sklearn if it is different from the three used earlier.\n - Train the supervised model on the entire training set.\n - Extract the feature importances using `'.feature_importances_'`.", "_____no_output_____" ] ], [ [ "# Import a supervised learning model that has 'feature_importances_'\nfrom sklearn.ensemble import AdaBoostClassifier\n\n# Train the supervised model on the training set \nmodel = AdaBoostClassifier(random_state=42).fit(X_train, y_train)\n\n# Extract the feature importances\nimportances = model.feature_importances_\n\n# Plot\nvs.feature_plot(importances, X_train, y_train)", "_____no_output_____" ] ], [ [ "### Question 7 - Extracting Feature Importance\n\nObserve the visualization created above which displays the five most relevant features for predicting if an individual makes at most or above \\$50,000. \n_How do these five features compare to the five features you discussed in **Question 6**? If you were close to the same answer, how does this visualization confirm your thoughts? If you were not close, why do you think these features are more relevant?_", "_____no_output_____" ] ], [ [ "# print top 10 features importances\ndef rank_features(features, scores, descending=True, n=10):\n \"\"\"\n sorts and cuts features by scores.\n :return: array of [feature name, score] tuples\n \"\"\"\n return sorted(\n [[f, s] for f, s in zip(features, scores) if s],\n key=lambda x: x[1],\n reverse=descending\n )[:n]\n\nrank_features(\n features.columns,\n importances\n)", "_____no_output_____" ], [ "# capital-loss and income have a positive correlation\nfeatures['capital-loss'].corr(income)", "_____no_output_____" ] ], [ [ "**Answer:**\n\nFrom the top 5 features selected by *AdaBoostClassifier* we got 4 hits (*age*, *capital-gain*, *hours-per-week* and *education-level*). That *capital-loss* has a such big influence is really surprising and by looking at the cell above, *income* and *capital-loss* are even positively correlated. Our top one guess *occupation* hasn't made into top 10 in the feature importances.\n\nThe visualization of the feature importances confirms our guess that our observation in the society is actually true that older and higher educated people with money (big *capital-loss* or *capital-gain*) would have higher salaries. These thoughts are now quantified and explained by the classfier.", "_____no_output_____" ], [ "### Feature Selection\nHow does a model perform if we only use a subset of all the available features in the data? With less features required to train, the expectation is that training and prediction time is much lower — at the cost of performance metrics. From the visualization above, we see that the top five most important features contribute more than half of the importance of **all** features present in the data. This hints that we can attempt to *reduce the feature space* and simplify the information required for the model to learn. The code cell below will use the same optimized model you found earlier, and train it on the same training set *with only the top five important features*. ", "_____no_output_____" ] ], [ [ "# Import functionality for cloning a model\nfrom sklearn.base import clone\n\n# Reduce the feature space\nX_train_reduced = X_train[X_train.columns.values[(np.argsort(importances)[::-1])[:5]]]\nX_test_reduced = X_test[X_test.columns.values[(np.argsort(importances)[::-1])[:5]]]\n\n# Train on the \"best\" model found from grid search earlier\nstart = time()\nclf = (clone(best_clf)).fit(X_train_reduced, y_train)\ntraining_time_reduced = time() - start\n\n# Make new predictions\nreduced_predictions = clf.predict(X_test_reduced)\n\n# Report scores from the final model using both versions of data\nprint \"Final Model trained on full data\\n------\"\nprint \"Accuracy on testing data: {:.4f}\".format(accuracy_score(y_test, best_predictions))\nprint \"F-score on testing data: {:.4f}\".format(fbeta_score(y_test, best_predictions, beta = 0.5))\nprint \"\\nFinal Model trained on reduced data\\n------\"\nprint \"Accuracy on testing data: {:.4f}\".format(accuracy_score(y_test, reduced_predictions))\nprint \"F-score on testing data: {:.4f}\".format(fbeta_score(y_test, reduced_predictions, beta = 0.5))", "Final Model trained on full data\n------\nAccuracy on testing data: 0.8514\nF-score on testing data: 0.7063\n\nFinal Model trained on reduced data\n------\nAccuracy on testing data: 0.8096\nF-score on testing data: 0.5983\n" ], [ "# compare scores\ndef relative_diff_pct(x, y):\n \"\"\"\n returns the relative difference between x and y in percent.\n \"\"\"\n return 100 * ((y - x) / x)\n\nprint('Relative Diff. of accuracy-scores: {0:.2f}%'.format(\n relative_diff_pct(.8514, .8096)\n))\nprint('Relative Diff. of f-scores: {0:.2f}%'.format(\n relative_diff_pct(.7063, .5983)\n))", "Relative Diff. of accuracy-scores: -4.91%\nRelative Diff. of f-scores: -15.29%\n" ], [ "# Train with full data\nstart = time()\nclf = (clone(best_clf)).fit(X_train, y_train)\ntraining_time_full = time() - start\n\nprint('Relative Diff. of training times: {0:.2f}%'.format(\n relative_diff_pct(training_time_reduced, training_time_full)\n))", "Relative Diff. of training times: 94.68%\n" ] ], [ [ "### Question 8 - Effects of Feature Selection\n*How does the final model's F-score and accuracy score on the reduced data using only five features compare to those same scores when all features are used?* \n*If training time was a factor, would you consider using the reduced data as your training set?*", "_____no_output_____" ], [ "**Answer:**", "_____no_output_____" ], [ "Both the accuracy and f-scores dropped down when using only the top 5 features. The f-scores have an **over 15% difference**, while the difference between the accuracy scores is at about 5%.\n\nThe training time with reduced data was more than **2 times faster** than the training time of the full data set. In some other scenarios when the training time or the computation power if of high priority, we would consider to use the reduced data. But since the difference between the f-scores are considerably large we would use the full training set for this problem.", "_____no_output_____" ], [ "## References\n\n* [Udacity - Machine Learning](https://classroom.udacity.com/courses/ud262)\n* [Laurad Hamilton - ML Cheat Sheet](http://www.lauradhamilton.com/machine-learning-algorithm-cheat-sheet)\n* [Scikit-Learn - ML Modules](http://scikit-learn.org/stable/modules/sgd.html)\n* [Scikit-Learn - AdaBoostClassifier](http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.AdaBoostClassifier.html)\n* [rcompton - Supervised Learning Superstitions](https://github.com/rcompton/ml_cheat_sheet)\n* [Wikipedia - Support Vector Machine](https://en.wikipedia.org/wiki/Support_vector_machine#Definition)\n* [Quora - SVM in layman's terms](https://www.quora.com/What-does-support-vector-machine-SVM-mean-in-laymans-terms)", "_____no_output_____" ], [ "## Reproduction Environment", "_____no_output_____" ] ], [ [ "import IPython\nprint IPython.sys_info()", "{'commit_hash': u'5c9c918',\n 'commit_source': 'installation',\n 'default_encoding': 'cp936',\n 'ipython_path': 'C:\\\\dev\\\\anaconda\\\\lib\\\\site-packages\\\\IPython',\n 'ipython_version': '5.1.0',\n 'os_name': 'nt',\n 'platform': 'Windows-7-6.1.7601-SP1',\n 'sys_executable': 'C:\\\\dev\\\\anaconda\\\\python.exe',\n 'sys_platform': 'win32',\n 'sys_version': '2.7.13 |Anaconda custom (32-bit)| (default, Dec 19 2016, 13:36:02) [MSC v.1500 32 bit (Intel)]'}\n" ], [ "!pip freeze", "alabaster==0.7.9\nanaconda-client==1.6.0\nanaconda-navigator==1.4.3\nargcomplete==1.0.0\nastroid==1.4.9\nastropy==1.3\nBabel==2.3.4\nbackports-abc==0.5\nbackports.shutil-get-terminal-size==1.0.0\nbackports.ssl-match-hostname==3.4.0.2\nbeautifulsoup4==4.5.3\nbitarray==0.8.1\nblaze==0.10.1\nbokeh==0.12.4\nboto==2.45.0\nBottleneck==1.2.0\ncdecimal==2.3\ncffi==1.9.1\nchardet==2.3.0\nchest==0.2.3\nclick==6.7\ncloudpickle==0.2.2\nclyent==1.2.2\ncolorama==0.3.7\ncomtypes==1.1.2\nconda==4.3.16\nconfigobj==5.0.6\nconfigparser==3.5.0\ncontextlib2==0.5.4\ncookies==2.2.1\ncryptography==1.7.1\ncycler==0.10.0\nCython==0.25.2\ncytoolz==0.8.2\ndask==0.13.0\ndatashape==0.5.4\ndecorator==4.0.11\ndill==0.2.5\ndocutils==0.13.1\nenum34==1.1.6\net-xmlfile==1.0.1\nfastcache==1.0.2\nFlask==0.12\nFlask-Cors==3.0.2\nfuncsigs==1.0.2\nfunctools32==3.2.3.post2\nfutures==3.0.5\ngevent==1.2.1\nglueviz==0.9.1\ngreenlet==0.4.11\ngrin==1.2.1\nh5py==2.6.0\nHeapDict==1.0.0\nidna==2.2\nimagesize==0.7.1\nipaddress==1.0.18\nipykernel==4.5.2\nipython==5.1.0\nipython-genutils==0.1.0\nipywidgets==5.2.2\nisort==4.2.5\nitsdangerous==0.24\njdcal==1.3\njedi==0.9.0\nJinja2==2.9.4\njsonschema==2.5.1\njupyter==1.0.0\njupyter-client==4.4.0\njupyter-console==5.0.0\njupyter-core==4.2.1\nlazy-object-proxy==1.2.2\nllvmlite==0.15.0\nlocket==0.2.0\nlxml==3.7.2\nMarkupSafe==0.23\nmatplotlib==2.0.0\nmenuinst==1.4.4\nmistune==0.7.3\nmpmath==0.19\nmultipledispatch==0.4.9\nnbconvert==4.2.0\nnbformat==4.2.0\nnetworkx==1.11\nnltk==3.2.2\nnose==1.3.7\nnotebook==4.4.1\nnumba==0.30.1+0.g8c1033f.dirty\nnumexpr==2.6.1\nnumpy==1.11.3\nnumpydoc==0.6.0\nodo==0.5.0\nopenpyxl==2.4.1\npandas==0.19.2\npartd==0.3.7\npath.py==0.0.0\npathlib2==2.2.0\npatsy==0.4.1\npep8==1.7.0\npickleshare==0.7.4\nPillow==4.0.0\nply==3.9\nprompt-toolkit==1.0.9\npsutil==5.0.1\npy==1.4.32\npyasn1==0.1.9\npycosat==0.6.1\npycparser==2.17\npycrypto==2.6.1\npycurl==7.43.0\npyflakes==1.5.0\npygame==1.9.3\nPygments==2.1.3\npylint==1.6.4\npymongo==3.3.0\npyOpenSSL==16.2.0\npyparsing==2.1.4\npytest==3.0.5\npython-dateutil==2.6.0\npytz==2016.10\npywin32==220\nPyYAML==3.12\npyzmq==16.0.2\nQtAwesome==0.4.3\nqtconsole==4.2.1\nQtPy==1.2.1\nrequests==2.12.4\nrequests-file==1.4.1\nresponses==0.5.1\nrope==0.9.4\nscandir==1.4\nscikit-image==0.12.3\nscikit-learn==0.18.1\nscipy==0.18.1\nseaborn==0.7.1\nsimplegeneric==0.8.1\nsingledispatch==3.4.0.3\nsix==1.10.0\nsnowballstemmer==1.2.1\nsockjs-tornado==1.0.3\nsphinx==1.5.1\nspyder==3.1.2\nSQLAlchemy==1.1.5\nstatsmodels==0.6.1\nsubprocess32==3.2.7\nsympy==1.0\ntables==3.2.2\ntoolz==0.8.2\ntornado==4.4.2\ntraitlets==4.3.1\nunicodecsv==0.14.1\nwcwidth==0.1.7\nWerkzeug==0.11.15\nwidgetsnbextension==1.2.6\nwin-unicode-console==0.5\nwrapt==1.10.8\nxlrd==1.0.0\nXlsxWriter==0.9.6\nxlwings==0.10.2\nxlwt==1.2.0\n" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Describing continuous variables using Probability Density Functions", "_____no_output_____" ] ], [ [ "import numpy as np\nimport matplotlib.pyplot as plt", "_____no_output_____" ], [ "data = np.random.normal(0.5, 0.1, 1000)", "_____no_output_____" ], [ "histogram = plt.hist(data, bins=10, range=(0.1, 1.5))", "_____no_output_____" ], [ "histogram = plt.hist(data, bins=20, range=(0.1, 1.5), density=True)", "_____no_output_____" ], [ "height = histogram[0][6].round(4)", "_____no_output_____" ], [ "x1 = histogram[1][6].round(4)", "_____no_output_____" ], [ "x2 = histogram[1][7].round(4)", "_____no_output_____" ], [ "3.24 * 0.07", "_____no_output_____" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# __General basic approach for applying Data Science.__\n- Collect Data.<br>\n- Extract features.<br>\n- Extract the target(label).<br>\n- Select the Estimator for learning.<br>\n- Tune the parameters.<br>\n- Fit the train data set.<br> \n- Test against testing_data_set.<br>\n- Check accuracy.<br>\n- Deploy to production.<br>\n- Write unit test cases for model.<br>\n- write live notification for model.<br>\n- Keep trainig the model with new data(fresh data).<br>", "_____no_output_____" ], [ "__Seaborn is a Python data visualization library based on matplotlib. It provides a high-level interface for drawing attractive and informative statistical graphics.__", "_____no_output_____" ] ], [ [ "#Import Seaborn\nimport seaborn as sns", "_____no_output_____" ] ], [ [ "__Seaborn comes with the iris data set , all we need is to load it. After loading we can do some spy things over data__\n", "_____no_output_____" ], [ "<img src=\"iris-machinelearning.png\" />", "_____no_output_____" ] ], [ [ "#Load iris data set from Sea born\niris = sns.load_dataset(\"iris\")\niris.head(4)", "_____no_output_____" ], [ "%matplotlib inline\nimport seaborn as sns; \nsns.set()\nsns.pairplot(iris, hue='species', size=3.5);", "_____no_output_____" ] ], [ [ "__Drop 'Species' feature from feature matrix, and look at the shape.__", "_____no_output_____" ] ], [ [ "iris.shape", "_____no_output_____" ], [ "#Perform basic EDA \niris.describe()", "_____no_output_____" ], [ "#Spy over how many outcomes are present in the Dataset\niris.species.unique()", "_____no_output_____" ] ], [ [ "__Time to split the iris dataset into Training:Tesing datset. Remember there is no standard approach fro this dividation even though we divide, Based on suggestions from ML/Data science leaders 70:30 approach is good.__", "_____no_output_____" ] ], [ [ "#Import train_test_split\nfrom sklearn.model_selection import train_test_split", "_____no_output_____" ], [ "#Split iris dataset into training and testing datset\ntrainIris , testIris = train_test_split(iris,test_size = 0.3)", "_____no_output_____" ], [ "#Look over training set\ntrainIris.head()", "_____no_output_____" ], [ "#Look over testing set\ntestIris.head()", "_____no_output_____" ] ], [ [ "__Testing dataset must not hold the target variable/outcomes, so that we can predict the outcome \nusing our trained regression model from trainig datset__ ", "_____no_output_____" ] ], [ [ "#Drop Species from testing dataset\ntestIris = testIris.drop(['species'],axis=1)", "_____no_output_____" ], [ "#Test set after dropping target/outcome column\ntestIris.head()", "_____no_output_____" ] ], [ [ "__As we have our training and testing datset ready, we have observed the Iris dataset,\nand have performed basic EDA over it. We can clearly see that this supervised learning problem.__", "_____no_output_____" ], [ "__TIme to pick some ML classifier, Train it , Test it__ &#9684;", "_____no_output_____" ], [ "__TO BE CONTINUED . . . .__", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Structural Transformation Notes\n\nBelow some brief notes on general equilibrium modeling of structural transformation. Some of the presentation illustrates and expands upon this short useful survey:\n\n> Matsuyama, K., 2008. Structural change. in Durlauf and Blume eds. *The new Palgrave dictionary of economics* 2, pp.\n\nThe note is organized around the following sections:\n\n- Market Clearing Neo-Classical Models\n - Push out of agriculture vs. Pull into Manufacturing\n - Pull in an Open Economy Model\n - Push in Closed Economy with Non-homothetic preferences\n- Productivity growth caused by structural change\n- Impediments to structural change\n- Spillovers, Complementarities, and Multiple Equilibria \n \n---\n", "_____no_output_____" ], [ "## Market Clearing Neo-Classical Models\n\nHere we seek to explain structural transformation and the accompanying movement of labor out of agriculture and into other sectors in a market-clearing neo-classical model with no market failures. In these model the marginal value product of labor is the same in both agriculture and manufacturing, hence there is no 'dualism' due to market frictions of one kind or another.\n\nMatsuyama builds a very simple neo-classical model with two sectors (in effect a specific factors model). \n\n**Production**\n\nSector $j \\in (1,2)$ is given by $Y_j = A_j \\cdot F_j(n_j)$. Let $j=1$ be the primary (agricultural) sector and $j=2$ the secondary (manufacturing) sector, where $A_j$ represents Total Factor Productivity (TFP) in each sector. We normalize the size of the population to 1. Then, if $n$ is the share (and total) labor used in sector 1 then $1-n$ is the amount left to be in sector two. \n\nMatsuyama refers to $F$ only as a generic increasing concave function, but for purposes of the graph examples, let's specify:\n\n$$\nA_1 F_1(n) = A_1 n^\\alpha \\\\\nA_2 F_2(1-n) = A_2 (1-n)^\\nu \n$$", "_____no_output_____" ], [ "Competitive markets will insure that labor is efficiently allocated across sectors such that the marginal value product of labor be equalized in each sector:\n \n$$\nA_1 F_1^\\prime (n) = p A_2 F_2^\\prime (1-n)\n$$", "_____no_output_____" ], [ "where $p=P_2/P_1$ is the relative price of sector 2 goods.", "_____no_output_____" ], [ "We can trace out a production possibility frontier (PPF) by simply plotting $\\left( A_1 F_1(n), A F_2(1-n) \\right)$ as we vary $n$ from 0 to 1. Below we draw two PPFs. One with $A_1=A_2=1$ and another afteragricultural TFP has improved to $A_1=1.2$ (we are setting $\\alpha=\\nu=1/2$)", "_____no_output_____" ] ], [ [ "import matplotlib.pyplot as plt\nimport numpy as np\nfrom scipy.optimize import fsolve", "_____no_output_____" ], [ "def F(n, a):\n return n ** a\n\ndef Fprime(n, a):\n return a* n ** (a-1)\n\n\ndef PPF(A1=1, A2=1, a1=0.5, a2=0.5, ax=None): \n if ax is None:\n ax = plt.gca()\n n = np.linspace(0,1,50)\n plt.plot( A1*F(n, a1), A2*F(1-n, a2) )\n plt.xlabel(r'$Y_1$'), plt.ylabel(r'$Y_2$')", "_____no_output_____" ], [ "fig, ax = plt.subplots(1)\nPPF(A1=1, A2=1, ax=ax)\nPPF(A1=1.3, A2=1, ax=ax)\nax.set_xlim(left=0), ax.set_ylim(bottom=0)\nax.set_aspect('equal')", "_____no_output_____" ] ], [ [ "### Push and/or Pull\n\nThe large literature on structural transformation often distinguishes between forces that 'Push' or 'Pull' labor out of agriculture. \n\n'Pull' could come about, for example, via an increase over time of the relative price of manufactures $p$, or an increase in relative TFP $A_2/A_1$. These have the effect of shifting out the demand for labor (the marginal value product or labor curve $p A_2 F_2^\\prime (1-n)$ relative to that in agriculture, leading to an incipient rise in the manufacturing wage that transforms the economy by attracting labor to that sector.\n\nAs demonstrated below, it's easy to illustrate Pull effects in this economy with standard homothetic preferences in either the open or closed economy.\n\nThe 'Push' argument is associated the idea that rises in agricultural productivity $A_1$ will mean that more food can be produced with fewer workers and therefore that labor can be 'released' from agriculture to other sectors. A related prevalent view is than an 'agricultural revolution' is often a prior and necessary condition for an industrial revolution. \n\nIt turns out one has to be a bit harder to get this effect to emerge in this standard neo-classical model because, as we've just argued, raising the relative productivity in a sector tends *increase* demand for labor unless the relative price of the good happens to simultaneously fall by enough to reverse that effect. As we'll show one way to engineer such an effect is to assume non-homothetic preferences to rise to a form of Engel's law or that manufactures have a higher income elasticity of demand, such that the agricultural terms of trade deteriorate against agriculture as incomes expand.", "_____no_output_____" ], [ "### The Open Economy with Homothetic Preferences\n\nIn the open-economy version of this model, the relative price of manufactured goods $p$ is set and fixed on world markets and hence does not change during the period of analysis. We can focus on the optimum production allocation without worrying about preferences, because agents first maximize the value of income at world prices and then choose optimum consumption baskets.\n\nWe can plot labor demand in each sector and solve for the equilibrium labor allocation $n^*$ that solves:\n\n$$\nA_1 F_1^\\prime (n) = p A_2 F_2^\\prime (1-n)\n$$", "_____no_output_____" ] ], [ [ "def weq(A1=1, A2=1, a1=0.5, a2=0.5, p=1):\n \n def foc(n):\n return p * A2 * Fprime(1-n, a2) - A1 * Fprime(n, a1) \n \n n = 0.75 # guess\n ne = fsolve(foc, n)[0]\n we = A1 * Fprime(ne, a1)\n return ne, we\n\n\ndef sfm(A1=1, A2=1, a1=0.5, a2=0.5, p=1, ax=None):\n if ax is None:\n ax = plt.gca()\n nn = np.linspace(0.01, 0.99, 100)\n ax.plot(nn, A1 * Fprime(nn, a1))\n ax.plot(nn, p * A2 * Fprime(1-nn, a2))\n ne, we = weq(A1, A2, a1, a2, p)\n ax.scatter(ne, we)\n ax.axhline(we, linestyle =':')\n ax.vlines(x = ne, ymin=0, ymax=we, linestyle =':')\n ax.set_ylim(0,2)\n ax.set_xlim(0,1)\n ax.set_xlabel(r'$n$')\n ax.text(0.8, 0.9*A1 *Fprime(1,a1), r'$L_1^d$')\n ax.text(0.1, 0.9*p*A2 *Fprime(0.999, a2), r'$L_2^d$')", "_____no_output_____" ] ], [ [ "**Pull: Impact of increase in relative price of manufactures in open Economy**\n\nExactly like a specific factors model diagram. A very similar diagram would depict effect of increase in sector 2 relative TFP $A_2/A_1$", "_____no_output_____" ] ], [ [ "sfm(p=1)\nsfm(p=1.5)", "_____no_output_____" ] ], [ [ "Exogenously driven increases in the relative productivity of manufactures drives this 'pull' effect. As Matsuyama explains, this is the sort of mechanism envisioned by Lewis (1954) although the Lewis model also has a form of dualism not captured here. In particular, we can see (from the diagram above) that in these models the expansion of one sector can be choked off by a rising wage. Lewis' assumption of \"unlimited supplies of labor\" in agriculture amounts to saying that there will be a very elastic supply of labor aso wages will not be fast to rise.\n", "_____no_output_____" ], [ "### Non-homothetic preferences", "_____no_output_____" ], [ "In a closed economy the relative price $p$ is determined endogenously by a tangency between the PPF and the representative indifference curve. As mentioned above, we'll need non-homothetic preferences in order to get the kind of 'Push' effect used in Gollin et al (2002):\n\nThe consumer has Stone-Geary preferences of the form:\n\n$$\nU(C_1, C_2) = \\beta \\cdot \\log(C_1 - \\gamma) + \\log(C_2)\n$$\n\nIf $\\gamma=0$ these would be standard homothetic preferences with linear income expansion paths (i.e. consumers maintain consumption ratio $C_2/C_1$ as income expands). When $\\gamma>0$ preferences are non-homothetic. We interpret $\\gamma$ as a minimum agricultural (food) consumption requirement. At low levels of income, all income is devoted to satisfying the requirement, but once that level is reached, remaining income will be spent in constant expenditure shares:", "_____no_output_____" ], [ "$$\nC_1 - \\gamma = \\frac{\\beta}{1+\\beta} \\frac{Y}{P_1} \\\\\nC_2 = \\frac{1}{1+\\beta} \\frac{Y}{P_2}\n$$", "_____no_output_____" ], [ "Define $p=P_2/P_1$ as the relative price of good 2, then $C_1$ and $C_2$ will be consumed proportional to each other following: \n\n$$\nC_1 =\\gamma + (\\beta p) C_2\n$$", "_____no_output_____" ], [ "Below is a plot of the optimum $C_2/C_1$ ratio for homothetic (blue, $\\gamma=0$) and non-homothetic (orange, $\\gamma=1$) preferences. In the latter case, after income reaches the level at which subsistence food needs are satisfied, the $C_2/C_1$ ratio (think of a ray from origin to a point on orange line) increases as incomes expand.", "_____no_output_____" ] ], [ [ "c1 = np.linspace(0,4,100)\ndef c2(c1, gam, beta, p):\n return (c1 - gam)/(beta * p)\n\nplt.plot(c1, c2(c1, 0, 0.5, 1))\nplt.plot(c1, c2(c1, 1, 0.5, 1))\nplt.ylim(0, 4), plt.xlim(0, 4)\nplt.xlabel(r'$C_1$'), plt.ylabel(r'$C_2$')\nplt.grid()\nplt.gca().set_aspect('equal')", "_____no_output_____" ] ], [ [ "### Closed Economy\n\n We're looking for a tangency between the PPF and the representative agent's indifference curve, equal to the common price ratio. This $MRS = p= MPT$ condition can be written:\n\n$$\n\\frac{1}{\\beta} \\frac{C_1 - \\gamma}{C_2} = p = \\frac{A_1 F_1^\\prime (n)}{A_2 F_2^\\prime (1-n)} \n$$\n\nUsing the fact that a closed economy must produce what it consumes we can substitute $C_1 = A_1 F_1(n)$ and $C_2 = A_2 F_2(1-n)$ and manipulate to obtain condition (2) in Matsuyama:\n\n$$\nF_1(n) - \\frac{\\beta F_2(1-n) F_1 ^\\prime (n)}{F_2 ^\\prime (1-n)} = \\frac{\\gamma}{A_1}\n$$\n", "_____no_output_____" ], [ "This implicitly defines $n$ as a decreasing function of $A_1$. Two things are striking here (both partly consequences of the Cobb-Douglas):\n\n1) $A_2$ plays no role. As $A_2$ rises $p$ falls by just enough to offset, leaving relative demand for labor unchanged.\n\n2) At any $\\gamma >0$ the agricultural labor share $n$ will fall with $A_1$. This is the Push effect.\n\nIn the plot below we plot this left hand side in red. The dashed line represents a value of $\\frac{\\gamma}{A_1}$. We can see that if $A_1$ rises (the dashed line moves down) the new equilibrium will involve less agricultural labor $n$.", "_____no_output_____" ] ], [ [ "def lhs(n, a1, a2, beta):\n F1 = F(n, a1)\n dF1 = Fprime(n, a1)\n F2 = F(1-n, a2)\n dF2 = Fprime(1-n, a2)\n return F1 - (beta*F2*dF1)/dF2 ", "_____no_output_____" ], [ "n = np.linspace(0.2,0.7,50)\nplt.plot(n, lhs(n, 0.5, 0.5, 0.5), color='r')\nplt.axhline(0);\nplt.axhline(0.5, linestyle='--')\nplt.xlabel(r'$n$');", "_____no_output_____" ] ], [ [ "We can solve for the closed economy equilibrium and plot things on a PPF diagram.", "_____no_output_____" ] ], [ [ "def neq(A1=1, a1=0.5, a2=0.5, beta=0.5, gamma= 0.5):\n '''Closed economy eqn from MRS=MPT'''\n def foc(n):\n return lhs(n, a1, a2, beta) - gamma/A1\n \n n = 0.7 # guess\n ne = fsolve(foc, n)[0]\n return ne", "_____no_output_____" ], [ "def plot_opt(A1, A2, a1, a2, beta, gamma):\n ne = neq(A1, a1, a2, beta, gamma)\n Y1 = A1 * F(ne, a1)\n Y2 = A2 * F(1-ne, a2)\n p = (1/beta) * (Y1-gamma)/Y2\n print(f'A1={A1}, n={ne:0.2f}, p={p:0.2f}')\n plt.scatter(Y1, Y2)\n ax =plt.gca()\n PPF(A1=A1, A2=A2, a1=a1, a2=a2, ax=ax)", "_____no_output_____" ] ], [ [ "Here we see structural transformation and a rise in the relative price of manufactures as TFP in agriculture increases:", "_____no_output_____" ] ], [ [ "plot_opt(1, 1, 0.5, 0.75, 1, 0.4)\nplot_opt(2, 1, 0.5, 0.75, 1, 0.4)\nplt.xlim(left=0)\nplt.ylim(bottom=0);", "A1=1, n=0.58, p=0.70\nA1=2, n=0.48, p=1.63\n" ] ], [ [ "### Productivity growth caused by Structural Change", "_____no_output_____" ], [ "Matsuyama adds dynamics to the sort of model described above by assuming that TFP growth in manufacturing might respond to \"the stock of experience accumulated [in that sector]... through learning-by-doing.\" He captures this by positing that TFP in the sector in period $t$ is given by:\n\n$$\nA_{2t} = A(Q_t)\n$$\n\nwhere $Q_t$ is the stock of experience which follows this law of motion:\n\n$$\n\\frac{dQ_t}{dt} = H(1-n_t)\n$$", "_____no_output_____" ], [ "where $H(\\cdot)$ is an increasing function. He assumes a critical level of manufacturing employment $(1-n_c)$ at which H(1-n_c)=0 (and negative at values below). With these assumptions and if $n<n_c$ so the critical threshold is surpassed, this plus the earlier closed economy model leads to a version of the **'Staple Theory of Growth'** whereby growth in the agricultural sector releases labor to manufacturing which over time generates productivity growth there. \n\nThe **Staples Thesis** of Growth (Innis, Watkins, others) emphasizes the role of the expansion of frontiers and the growth of traditional 'staple' products such as wheat in a resource-rich economies. This has been used to explain growth and development in regions of the world including Canada, northern United States, and to some extent Australia, New Zealand and Argentina. These countries benefited from an extraordinary rise in commodity prices in the late 19th century which, by many accounts led to the growth of supporting manufacturing industries, in a manner similar to that captured by this account.\n\nMatsuyama points out however that a variation on the same model can explain a version of the **resource curse**. For example in the small open economy the equilibrium condition is as above:\n\n$$\nA_1 F_1^\\prime (n_t) = p A(Q_t) F_2^\\prime (1-n_t)\n$$\n\nAs discussed before, raising $A_1$ implies higher $n_t$ for any level of $Q_t$. If the economy is already passed the threshold $H(1-n_t)>0$ which leads to growth in manufacturing and hence pull out of agriculture, and the effect is self-reinforcing.\n\nHowever if below the threshold (i.e. $n_t>n_c$) and hence $H(1-n_t)<0$ anything that stimulates agriculture, be it productivity growth or a rise in the relative price of agricultural products, will lead to productivity *decline* in industry and a steady increase in agricultural employment $n_t$.\n\nThis is a version of the so called *staple-trap,* *resource curse,* or *dutch disease*. Matsuyama (1992) models this type of possibility in richer detail.", "_____no_output_____" ], [ "### Impediments to structural change", "_____no_output_____" ], [ "Many studies have modeled impediments to the re-allocation of labor across sectors. \n\nThe most famous of these is the Lewis model which assumes workers in the agricultural sector earn an average product of labor there, leading to overemployment in the sector. We'll see several models that try to make sense of this, for example models that attribute this to land property rights insecurity. \n\nHayashi and Prescott (2008) argue that agricultural employment in Japan remained virtually constant at 14 million persons from 1885 to WWII (see image below) despite significant expansion in industrial technology capacity in that period. They argue that labor could not easily leave agriculture due to patriarchal customs and norms that made it hard for children of agricultural households to leave the parental household). Using a calibrated general equilibrium model, the authors argue that the lifting of that barrier in the post-war era (due to various factors including legislation surrounding deep land reforms) explains a good part of the post-war Japanese miracle. ", "_____no_output_____" ], [ "<img src=\"../media/Japan_agemp.png\" alt=\"Japan Ag employment\" class=\"bg-primary\" width=\"600px\">", "_____no_output_____" ], [ "There are many other models with barriers or frictions in the structural transformation. One famous model that received a lot of attentiun is the Harris and Todaro (1970) model. See the separate notebook on this model. ", "_____no_output_____" ], [ "### Spillovers, Complementarities, and Multiple Equilibria \n\nComing soon.", "_____no_output_____" ], [ "## References Cited\n\nHarris, J.R., Todaro, M.P., 1970. Migration, unemployment and development: a two-sector analysis. *The American economic review* 126–142.\n\nHayashi, F., Prescott, E.C., 2008. The depressing effect of agricultural institutions on the prewar Japanese economy. *Journal of political Economy* 116, 573–632.\n\nLewis, W.A., 1954. Economic development with unlimited supplies of labour. *The Manchester School* 22, 139–191.\n\nMatsuyama, K., 2008. Structural change. *The new Palgrave dictionary of economics* 2.\n\nMatsuyama, K., 1992. Agricultural productivity, comparative advantage, and economic growth. *Journal of economic theory* 58, 317–334.\n\n", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Advent-of-Code-2019\n\n## About Advent of Code\n\n Advent of Code is an Advent calendar of small programming puzzles for a variety of skill sets and skill levels that can be solved in any programming language you like. People use them as a speed contest, interview prep, company training, university coursework, practice problems, or to challenge each other.\n\n## Websites\n\n https://adventofcode.com/\n ServiceNow Leaderboard: https://adventofcode.com/2019/leaderboard/private/view/109388]\n PySlackers Leaderboard: https://adventofcode.com/2019/leaderboard/private/view/52704]\n\n## 2019 theme\n\n Santa has become stranded at the edge of the Solar System while delivering presents to other planets! \n To accurately calculate his position in space, safely align his warp drive, and \n return to Earth in time to save Christmas, he needs you to bring him measurements from fifty stars.\n\n Collect stars by solving puzzles. Two puzzles will be made available on each day in the Advent calendar; \n the second puzzle is unlocked when you complete the first. \n Each puzzle grants one star. Good luck!\n", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Feature Extraction\nIn machine learning, feature extraction aims to compute values (features) from images, intended to be informative and non-redundant, facilitating the subsequent learning and generalization steps, and in some cases leading to better human interpretations. These features may be handcrafted (manually computed features based on a-priori information ) or convolutional features (detect patterns within the data without the prior definition of features or characteristics):\n\n## Handcrafted: \n- Histogram features: Statistical moments extracted from the histogram describe image characteristics and provide a quantitative analysis of the image intensity distribution (entropy, intensity mean, standard deviation, skewness, kurtosis, and values at 0, 10, 50 (median) and 90 percentiles).\n<img src=\"../Figures/histogram_sample.jpg\" alt=\"Drawing\" style=\"width: 500px;\"/>\n- The gradient examines directional changes in the image gray level, and can extract relevant information (such as edges) within an image. Moments extracted from the image gradient are more robust to acquisition conditions, such as contrast variation, and properties of the acquisition equipment . Ten features were extracted from the gray level and morphological gradients (five from each): intensity mean, standard deviation, skewness, kurtosis and percentage of non-zero values.\n<img src=\"../Figures/gradient.png\" alt=\"Drawing\" style=\"width: 300px;\"/>\n- Local binary pattern (LBP) is a texture spectrum model that may be used to identify patterns in an image. The LBP histogram comprises the frequency of occurrence of different patterns within an image. Ten features were extracted from the LBP by using a 10-bin LBP histogram.\n<img src=\"../Figures/lbp.png\" alt=\"Drawing\" style=\"width: 400px;\"/>\n- The Haar wavelet is a multi-resolution technique that transforms images into a domain where both spatial and frequency information is present. Features separately extracted from each sub-image present desired scale-dependent properties. When considering two decomposition levels, eight sub-images are generated. The mean value within each sub-image were computed and used as features (total of eight features). <img src=\"../Figures/wavelet-haar.png\" alt=\"Drawing\" style=\"width: 400px;\"/>\n\n\n## Convolutional features\nComputed by using a very deep convolutional network (VGG16) with pre-trained imagenet weights. For each MR volume, the convolutional features were computed in the central 2D axial, sagittal and coronal slices. For each of these three views, 25,088 convolutional features were computed and combined. \n<img src=\"../Figures/viz_initial_layers.png\" alt=\"Drawing\" style=\"width: 700px;\"/>\n\nBesides the image and patient information (MR vendor, magnetic field, age, gender), a total of 75,300 features were extracted and combined for each image: 8 features from the image histogram, 10 features from the image gradient, 10 features from the LBP histogra, 8 features from the Haar wavelet subimages and 75,264 convolutional features. \n\n<img src=\"../Figures/data.png\" alt=\"Drawing\" style=\"width: 800px;\"/>\n", "_____no_output_____" ], [ "## Reading the data file containing patients information and features", "_____no_output_____" ] ], [ [ "## data: vendor; magnetic field; age; gender; feats (65300) \n # vendor: ge -> 10; philips -> 11; siemens -> 12\n # gender: female -> 10; male -> 11\n # feats: fs1 - histogram (8); fs2 - gradient (10); fs3 - lbp (10); fs4 - haar (8); fs5 - convolutional (75264)\nimport numpy as np\n\ndata = np.load('../Data/feats_cc359.npy.zip')['feats_cc359']\nprint '#samples, #info: ',data.shape\n\nprint 'patients age:', data[:,2]", "#samples, #info: (359, 75304)\npatients age: [ 55. 56. 63. 67. 62. 63. 62. 60. 69. 69. 49. 43. 66. 62. 44.\n 55. 50. 41. 57. 65. 48. 43. 43. 65. 51. 65. 41. 63. 51. 42.\n 65. 44. 67. 43. 49. 49. 41. 41. 41. 55. 61. 67. 58. 36. 49.\n 42. 54. 53. 43. 45. 44. 51. 39. 46. 44. 39. 61. 64. 55. 29.\n 55. 52. 53. 49. 42. 46. 57. 49. 56. 80. 56. 31. 49. 53. 41.\n 44. 43. 42. 47. 55. 67. 64. 66. 46. 71. 42. 42. 51. 46. 37.\n 51. 60. 60. 45. 43. 36. 45. 48. 51. 60. 52. 50. 56. 34. 52.\n 41. 38. 51. 57. 44. 51. 60. 45. 60. 49. 50. 49. 49. 53. 58.\n 61. 53. 58. 57. 65. 54. 58. 54. 54. 57. 41. 71. 64. 54. 52.\n 64. 65. 56. 56. 57. 53. 59. 55. 43. 62. 38. 58. 58. 53. 47.\n 71. 51. 52. 55. 57. 55. 54. 53. 54. 57. 55. 39. 64. 54. 44.\n 39. 55. 38. 44. 55. 51. 52. 51. 41. 50. 45. 57. 55. 53. 39.\n 36. 60. 58. 60. 52. 62. 63. 55. 66. 63. 59. 51. 64. 67. 55.\n 48. 54. 59. 64. 37. 58. 54. 56. 49. 64. 58. 58. 56. 57. 60.\n 65. 67. 65. 51. 46. 56. 45. 61. 49. 61. 61. 56. 56. 62. 47.\n 64. 58. 66. 55. 60. 56. 56. 52. 57. 61. 50. 59. 58. 56. 60.\n 60. 55. 58. 56. 48. 60. 53. 51. 58. 60. 41. 44. 58. 57. 57.\n 42. 57. 55. 56. 51. 59. 54. 51. 59. 56. 57. 60. 57. 59. 59.\n 60. 44. 56. 47. 50. 52. 42. 61. 49. 57. 56. 57. 44. 41. 56.\n 55. 53. 56. 59. 56. 48. 60. 50. 51. 55. 57. 57. 58. 37. 58.\n 53. 56. 53. 57. 59. 52. 57. 52. 42. 51. 50. 49. 34. 54. 59.\n 57. 45. 52. 45. 57. 51. 59. 56. 56. 55. 55. 50. 56. 58. 57.\n 61. 53. 41. 45. 54. 53. 59. 51. 59. 49. 60. 50. 54. 43. 53.\n 43. 52. 60. 60. 62. 52. 55. 58. 61. 55. 61. 55. 52. 58.]\n" ] ], [ [ "## Activities List\n\n - Print the vendor for all the images\n - Identify images acquired from patients with age > 30\n - Identify images acquired using GE vendor scanner\n - Access this simple demo of a convnet trained on MNIST dataset: https://transcranial.github.io/keras-js/#/mnist-cnn\n - draw a number (between 0-9)\n - Visualize classification and intermediate outputs at each layer", "_____no_output_____" ], [ "## References\n- Histogram basics: https://docs.opencv.org/3.1.0/d1/db7/tutorial_py_histogram_begins.html\n- Local binary patterns: http://scikit-image.org/docs/dev/auto_examples/features_detection/plot_local_binary_pattern.html\n- Wavelet haar to MRI: https://www.researchgate.net/figure/The-procedures-of-3-level-2D-DWT-a-normal-brain-MRI-b-level-3-wavelet-coefficients_258104202\n- Extracting convolutional features using Vgg16: https://keras.io/applications/", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Imports", "_____no_output_____" ] ], [ [ "import pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nfrom sklearn import ensemble\nfrom sklearn import metrics\nfrom io import StringIO\nfrom csv import writer", "_____no_output_____" ] ], [ [ "# Read in csv files", "_____no_output_____" ] ], [ [ "matches = pd.read_csv('../csv/matches.csv')\nplayers = pd.read_csv('../csv/players.csv')\nhero_names = pd.read_json('../json/heroes.json')\ncluster_regions = pd.read_csv('./Data/cluster_regions.csv')", "_____no_output_____" ], [ "matches", "_____no_output_____" ], [ "players.head()", "_____no_output_____" ], [ "hero_names.head()", "_____no_output_____" ] ], [ [ "# Data info\n", "_____no_output_____" ], [ "## Hero Info\nMost and least popular heroes", "_____no_output_____" ] ], [ [ "num_heroes = len(hero_names)\nplt.hist(players['hero_id'], num_heroes)\nplt.show()", "_____no_output_____" ], [ "hero_counts = players['hero_id'].value_counts().rename_axis('hero_id').reset_index(name='num_matches')\npd.merge(hero_counts, hero_names, left_on='hero_id', right_on='id')", "_____no_output_____" ] ], [ [ "## Server Info\nWhere the most and least games are played", "_____no_output_____" ] ], [ [ "plt.hist(matches['cluster'], bins=np.arange(matches['cluster'].min(), matches['cluster'].max()+1))\nplt.show()", "_____no_output_____" ], [ "cluster_counts = matches['cluster'].value_counts().rename_axis('cluster').reset_index(name='num_matches')\npd.merge(cluster_counts, cluster_regions, on='cluster')", "_____no_output_____" ], [ "short_players = players.iloc[:, :11]\nshort_players.insert(short_players.shape[1], 'win', value=False)\nshort_players", "_____no_output_____" ], [ "# short_matches = matches.iloc[:1000]\nfor index, row in matches.iterrows():\n offset = 10 * index\n short_players.at[0 + offset, 'win'] = row.radiant_win\n short_players.at[1 + offset, 'win'] = row.radiant_win\n short_players.at[2 + offset, 'win'] = row.radiant_win\n short_players.at[3 + offset, 'win'] = row.radiant_win\n short_players.at[4 + offset, 'win'] = row.radiant_win\n\n short_players.at[5 + offset, 'win'] = not row.radiant_win\n short_players.at[6 + offset, 'win'] = not row.radiant_win\n short_players.at[7 + offset, 'win'] = not row.radiant_win\n short_players.at[8 + offset, 'win'] = not row.radiant_win\n short_players.at[9 + offset, 'win'] = not row.radiant_win\n # print(index)\n\nshort_players.head(20)", "_____no_output_____" ], [ "short_players.tail()", "_____no_output_____" ], [ "hero_match_wins = short_players.groupby(by='hero_id').sum()['win']\nhero_match_count = players['hero_id'].value_counts().rename_axis('hero_id').reset_index(name='total_matches')\nhero_match_count = hero_match_count.merge(hero_match_wins, on='hero_id')\nhero_match_count", "_____no_output_____" ], [ "hero_match_count['win_percent'] = hero_match_count['win'] / hero_match_count['total_matches'] * 100\nhero_match_count", "_____no_output_____" ], [ "hero_match_count = pd.merge(hero_match_count, hero_names, left_on='hero_id', right_on='id')\nhero_match_count", "_____no_output_____" ], [ "fig_size = plt.rcParams[\"figure.figsize\"]\nfig_size[0] = 10\nfig_size[1] = 8\nplt.rcParams[\"figure.figsize\"] = fig_size\n\n\nx = hero_match_count['total_matches']\ny = hero_match_count['win_percent']\nlabels = hero_match_count['localized_name']\n\n\nplt.scatter(x, y)\nplt.title('Win Percent vs Number of Games Played', fontsize=18)\nplt.xlabel('Number of Games Played', fontsize=18)\nplt.ylabel('Percent Won', fontsize=18)\n\nfor i, label in enumerate(labels):\n plt.annotate(label, (x[i], y[i]))", "_____no_output_____" ] ], [ [ "# Data cleaning\nWe start with an empty list of DataFrams and add to it as we create DataFrames of bad match ids. In the end we combine all the DataFrames and remove their match ids from the Matches DataFrame.", "_____no_output_____" ] ], [ [ "dfs_bad_matches = []", "_____no_output_____" ] ], [ [ "## Abandons\nremove games were a player has abandoned the match", "_____no_output_____" ] ], [ [ "abandoned_matches = players[players.leaver_status > 1][['match_id']]\nabandoned_matches = abandoned_matches.drop_duplicates().reset_index(drop=True)\ndfs_bad_matches.append(abandoned_matches)\nabandoned_matches", "_____no_output_____" ] ], [ [ "## Missing Hero id\nremove games where a player is not assigned a hero id, but didnt get flaged for an abandon", "_____no_output_____" ] ], [ [ "player_no_hero = players[players.hero_id == 0][['match_id']].reset_index(drop=True)\ndfs_bad_matches.append(player_no_hero)\nplayer_no_hero", "_____no_output_____" ] ], [ [ "## Wrong Game Mode\nremove games not played in \"Ranked All Pick\" (22)", "_____no_output_____" ] ], [ [ "wrong_mode = matches[matches.game_mode != 22].reset_index()[['match_id']]\ndfs_bad_matches.append(wrong_mode)\nwrong_mode", "_____no_output_____" ] ], [ [ "## Game length (short)\nremove games we deem too short (< 15 min)", "_____no_output_____" ] ], [ [ "short_length = 15 * 60\nshort_matches = matches[matches.duration < short_length].reset_index()[['match_id']]\ndfs_bad_matches.append(short_matches)\nshort_matches", "_____no_output_____" ] ], [ [ "## Game length (long)\nNext we want to get matches with a too long duration (>90 min)", "_____no_output_____" ] ], [ [ "long_length = 90 * 60\nlong_matches = matches[matches.duration > long_length].reset_index()[['match_id']]\ndfs_bad_matches.append(long_matches)\nlong_matches", "_____no_output_____" ] ], [ [ "## Combine all our lists of bad matches\ncombine matches and create a filtered match dataframe with only good matches", "_____no_output_____" ] ], [ [ "bad_match_ids = pd.concat(dfs_bad_matches, ignore_index=True).drop_duplicates()\nbad_match_ids", "_____no_output_____" ] ], [ [ "## Remove bad matches", "_____no_output_____" ] ], [ [ "filtered_matches = matches[~matches['match_id'].isin(bad_match_ids['match_id'])]\nfiltered_matches.info()", "<class 'pandas.core.frame.DataFrame'>\nInt64Index: 115823 entries, 0 to 145324\nData columns (total 22 columns):\n # Column Non-Null Count Dtype\n--- ------ -------------- -----\n 0 radiant_win 115823 non-null bool \n 1 duration 115823 non-null int64\n 2 pre_game_duration 115823 non-null int64\n 3 start_time 115823 non-null int64\n 4 match_id 115823 non-null int64\n 5 match_seq_num 115823 non-null int64\n 6 tower_status_radiant 115823 non-null int64\n 7 tower_status_dire 115823 non-null int64\n 8 barracks_status_radiant 115823 non-null int64\n 9 barracks_status_dire 115823 non-null int64\n 10 cluster 115823 non-null int64\n 11 first_blood_time 115823 non-null int64\n 12 lobby_type 115823 non-null int64\n 13 human_players 115823 non-null int64\n 14 leagueid 115823 non-null int64\n 15 positive_votes 115823 non-null int64\n 16 negative_votes 115823 non-null int64\n 17 game_mode 115823 non-null int64\n 18 flags 115823 non-null int64\n 19 engine 115823 non-null int64\n 20 radiant_score 115823 non-null int64\n 21 dire_score 115823 non-null int64\ndtypes: bool(1), int64(21)\nmemory usage: 19.6 MB\n" ] ], [ [ "## Remove duplicate matches", "_____no_output_____" ] ], [ [ "filtered_matches = filtered_matches.drop_duplicates(subset=['match_id'])\nfiltered_matches.info()", "<class 'pandas.core.frame.DataFrame'>\nInt64Index: 114686 entries, 0 to 145324\nData columns (total 22 columns):\n # Column Non-Null Count Dtype\n--- ------ -------------- -----\n 0 radiant_win 114686 non-null bool \n 1 duration 114686 non-null int64\n 2 pre_game_duration 114686 non-null int64\n 3 start_time 114686 non-null int64\n 4 match_id 114686 non-null int64\n 5 match_seq_num 114686 non-null int64\n 6 tower_status_radiant 114686 non-null int64\n 7 tower_status_dire 114686 non-null int64\n 8 barracks_status_radiant 114686 non-null int64\n 9 barracks_status_dire 114686 non-null int64\n 10 cluster 114686 non-null int64\n 11 first_blood_time 114686 non-null int64\n 12 lobby_type 114686 non-null int64\n 13 human_players 114686 non-null int64\n 14 leagueid 114686 non-null int64\n 15 positive_votes 114686 non-null int64\n 16 negative_votes 114686 non-null int64\n 17 game_mode 114686 non-null int64\n 18 flags 114686 non-null int64\n 19 engine 114686 non-null int64\n 20 radiant_score 114686 non-null int64\n 21 dire_score 114686 non-null int64\ndtypes: bool(1), int64(21)\nmemory usage: 19.4 MB\n" ], [ "filtered_players = players[~players['match_id'].isin(bad_match_ids['match_id'])]\nfiltered_players.info()", "<class 'pandas.core.frame.DataFrame'>\nInt64Index: 1158370 entries, 0 to 1453439\nData columns (total 31 columns):\n # Column Non-Null Count Dtype \n--- ------ -------------- ----- \n 0 account_id 1158308 non-null float64\n 1 player_slot 1158370 non-null int64 \n 2 hero_id 1158370 non-null int64 \n 3 item_0 1158370 non-null int64 \n 4 item_1 1158370 non-null int64 \n 5 item_2 1158370 non-null int64 \n 6 item_3 1158370 non-null int64 \n 7 item_4 1158370 non-null int64 \n 8 item_5 1158370 non-null int64 \n 9 backpack_0 1158370 non-null int64 \n 10 backpack_1 1158370 non-null int64 \n 11 backpack_2 1158370 non-null int64 \n 12 item_neutral 1158370 non-null int64 \n 13 kills 1158370 non-null int64 \n 14 deaths 1158370 non-null int64 \n 15 assists 1158370 non-null int64 \n 16 leaver_status 1158308 non-null float64\n 17 last_hits 1158370 non-null int64 \n 18 denies 1158370 non-null int64 \n 19 gold_per_min 1158370 non-null int64 \n 20 xp_per_min 1158370 non-null int64 \n 21 level 1158370 non-null int64 \n 22 hero_damage 1158370 non-null int64 \n 23 tower_damage 1158370 non-null int64 \n 24 hero_healing 1158370 non-null int64 \n 25 gold 1158370 non-null int64 \n 26 gold_spent 1158370 non-null int64 \n 27 scaled_hero_damage 1158370 non-null int64 \n 28 scaled_tower_damage 1158370 non-null int64 \n 29 scaled_hero_healing 1158370 non-null int64 \n 30 match_id 1158370 non-null int64 \ndtypes: float64(2), int64(29)\nmemory usage: 282.8 MB\n" ], [ "filtered_players = filtered_players.drop_duplicates(subset=['match_id', 'player_slot'])\nfiltered_players.info()", "<class 'pandas.core.frame.DataFrame'>\nInt64Index: 1146860 entries, 0 to 1453439\nData columns (total 31 columns):\n # Column Non-Null Count Dtype \n--- ------ -------------- ----- \n 0 account_id 1146801 non-null float64\n 1 player_slot 1146860 non-null int64 \n 2 hero_id 1146860 non-null int64 \n 3 item_0 1146860 non-null int64 \n 4 item_1 1146860 non-null int64 \n 5 item_2 1146860 non-null int64 \n 6 item_3 1146860 non-null int64 \n 7 item_4 1146860 non-null int64 \n 8 item_5 1146860 non-null int64 \n 9 backpack_0 1146860 non-null int64 \n 10 backpack_1 1146860 non-null int64 \n 11 backpack_2 1146860 non-null int64 \n 12 item_neutral 1146860 non-null int64 \n 13 kills 1146860 non-null int64 \n 14 deaths 1146860 non-null int64 \n 15 assists 1146860 non-null int64 \n 16 leaver_status 1146801 non-null float64\n 17 last_hits 1146860 non-null int64 \n 18 denies 1146860 non-null int64 \n 19 gold_per_min 1146860 non-null int64 \n 20 xp_per_min 1146860 non-null int64 \n 21 level 1146860 non-null int64 \n 22 hero_damage 1146860 non-null int64 \n 23 tower_damage 1146860 non-null int64 \n 24 hero_healing 1146860 non-null int64 \n 25 gold 1146860 non-null int64 \n 26 gold_spent 1146860 non-null int64 \n 27 scaled_hero_damage 1146860 non-null int64 \n 28 scaled_tower_damage 1146860 non-null int64 \n 29 scaled_hero_healing 1146860 non-null int64 \n 30 match_id 1146860 non-null int64 \ndtypes: float64(2), int64(29)\nmemory usage: 280.0 MB\n" ], [ "filtered_players.to_csv('../csv/players_filtered.csv', index=False)", "_____no_output_____" ], [ "filtered_matches.to_csv('../csv/matches_filtered.csv', index=False)", "_____no_output_____" ] ], [ [ "# Convert our match list\n\nConvert our match list to the form of :\nr_1, r_2, r_3, r_4, r_5, d_1, d_2, d_3, d_4, d_5, r_win\n", "_____no_output_____" ] ], [ [ "r_names = []\nd_names = []\nfor slot in range(1, 6):\n r_name = 'r_' + str(slot)\n d_name = 'd_' + str(slot)\n r_names.append(r_name)\n d_names.append(d_name)\n\ncolumns = (r_names + d_names + ['r_win'])\n\nnew_row = [-1] * (5 + 5 + 1)\n\n# test_players = players.iloc[:500, :]\n# test_matches = matches.iloc[:50, :]\n\ncolumns = (r_names + d_names + ['r_win'])\nnew_match_list = pd.DataFrame(data=None, columns=columns)\n\noutput = StringIO()\ncsv_writer = writer(output)\nfor index, row in filtered_matches.iterrows():\n match_players = players.loc[players['match_id'] == row['match_id']]\n\n if match_players.shape[0] == 10:\n\n new_row[0] = match_players[match_players['player_slot'] == 0].iloc[0]['hero_id']\n new_row[1] = match_players[match_players['player_slot'] == 1].iloc[0]['hero_id']\n new_row[2] = match_players[match_players['player_slot'] == 2].iloc[0]['hero_id']\n new_row[3] = match_players[match_players['player_slot'] == 3].iloc[0]['hero_id']\n new_row[4] = match_players[match_players['player_slot'] == 4].iloc[0]['hero_id']\n\n new_row[5] = match_players[match_players['player_slot'] == 128].iloc[0]['hero_id']\n new_row[6] = match_players[match_players['player_slot'] == 129].iloc[0]['hero_id']\n new_row[7] = match_players[match_players['player_slot'] == 130].iloc[0]['hero_id']\n new_row[8] = match_players[match_players['player_slot'] == 131].iloc[0]['hero_id']\n new_row[9] = match_players[match_players['player_slot'] == 132].iloc[0]['hero_id']\n\n # add if radiant win\n new_row[10] = row['radiant_win']\n \n # append new row onto match list\n csv_writer.writerow(new_row)\n\n # gives us some idea about how far into the data we are\n if index % 1000 == 0:\n print(index / 1000)\n\noutput.seek(0)\nnew_match_list = pd.read_csv(output, names=columns)\nnew_match_list.to_csv('./Data_new/new_match_list_filtered_ids.csv', index=False)", "_____no_output_____" ] ], [ [ "# Stats", "_____no_output_____" ] ], [ [ "players", "_____no_output_____" ], [ "player_stats = players.drop(columns=['account_id', 'match_id', 'leaver_status'])", "_____no_output_____" ], [ "player_stats_short = player_stats.drop(columns=['item_0','item_1','item_2','item_3','item_4','item_5','backpack_0','backpack_1','backpack_2','item_neutral', 'player_slot']).groupby(['hero_id']).mean()\nplayer_stats_short", "_____no_output_____" ], [ "player_stats_short.reset_index().to_json('../json/hero_stats.json', orient='records', indent=4)", "_____no_output_____" ], [ "player_items = player_stats.melt(id_vars=['hero_id'], value_vars=['item_0', 'item_1', 'item_2', 'item_3', 'item_4', 'item_5'], var_name='item_slot', value_name='item_id')\nplayer_items", "_____no_output_____" ], [ "item_counts = player_items.groupby(['hero_id']).item_id.value_counts().reset_index(name=\"count\")\nitem_counts", "_____no_output_____" ], [ "j = (item_counts.groupby(['hero_id'], as_index=True)\n .apply(lambda x: x[['item_id','count']].to_dict('r'))\n .reset_index()\n .rename(columns={0:'item_data'})\n .to_json('../json/item_stats.json', orient='records', indent=4))", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ] ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ [ [ "from enchant.tokenize import get_tokenizer\nimport enchant\n\ndic = enchant.Dict(\"en_US\")\ntknzr = get_tokenizer(\"en_US\")\n\nimport re\n\n# di\n# text = \"this's a sent tokenize, test. this is sent two. is this sent three? sent 4 is cool! Now it's your turn.\"\n# sent_tokenize_list = sent_tokenize(text)\n# print sent_tokenize_list\ndic.check(\".\")\n\" \".join(re.findall(r\"[\\w']+|[.,!?;]\", \"Hello, I'm a string!\"))", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "#Д\n#\n\nn = int(input('Количество вводимых значений: '))\n\nx = {}\n\nfor i in range(1, n + 1):\n \n y = []\n a = int(input('Значение: '))\n \n for j in range(2, a + 1):\n \n k = 0\n \n for d in range(1, j+1):\n \n if j % d == 0: k += 1\n \n if k == 2: y.append(d)\n \n x[i] = y\n \nfor key, value in x.items():\n print(\"{0}: {1}\".format(key,value))", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "data = pd.read_csv(\"/Users/kimjeongseob/Desktop/Study/0.Project/3. Machine Learning Practice/2. Football/dataset_football.csv\")\n", "_____no_output_____" ], [ "data = data.drop(columns = 'Unnamed: 0')", "_____no_output_____" ], [ "data_origin = data.copy()", "_____no_output_____" ], [ "data", "_____no_output_____" ], [ "data.follower = data.follower + 10\ndata.follower = np.log(data.follower)", "_____no_output_____" ], [ "# data = data.drop(columns='player_name')", "_____no_output_____" ], [ "col_name = list(data.columns)\ncol_name.remove('player_name')", "_____no_output_____" ], [ "# from patsy import *", "_____no_output_____" ], [ "# col_names = [\"scale({})\".format(name) for name in col_name]", "_____no_output_____" ], [ "# col_names", "_____no_output_____" ], [ "# data_scaled = dmatrix(\"0+\"+\"+\".join(col_names),data=data)", "_____no_output_____" ], [ "# data_scaled = pd.DataFrame(data_scaled,columns = [col_name])", "_____no_output_____" ], [ "# data_scaled", "_____no_output_____" ], [ "# from sklearn.preprocessing import MinMaxScaler\n\n# scaler = MinMaxScaler()\n# scaler.fit(data)\n# scaled = scaler.transform(data)", "_____no_output_____" ], [ "# from sklearn.preprocessing import StandardScaler\n\n# scaler = StandardScaler()\n# scaler.fit(data)\n# data_scaled = scaler.transform(data)", "_____no_output_____" ], [ "from sklearn.neighbors import KNeighborsClassifier", "_____no_output_____" ], [ "classifier = KNeighborsClassifier(n_neighbors = 100)", "_____no_output_____" ], [ "data_X = data.drop(columns='value')", "_____no_output_____" ], [ "data_y = data.value", "_____no_output_____" ], [ "data_y = map(int,data_y)", "_____no_output_____" ], [ "data_y = list(data_y)", "_____no_output_____" ], [ "# data_y", "_____no_output_____" ], [ "data_X", "_____no_output_____" ], [ "from sklearn.model_selection import train_test_split\n\nX_train, X_test, y_train, y_test = train_test_split(data_X, data_y, test_size=0.4, random_state=0)", "_____no_output_____" ], [ "X_train.shape", "_____no_output_____" ], [ "X_test.shape", "_____no_output_____" ], [ "len(y_train)", "_____no_output_____" ], [ "len(y_test)", "_____no_output_____" ], [ "classifier.fit(X_train.drop(columns='player_name'), y_train)", "_____no_output_____" ], [ "X_train", "_____no_output_____" ], [ "len(y_train)", "_____no_output_____" ], [ "print(classifier.score(X_test.drop(columns='player_name'),y_test))", "0.08743169398907104\n" ], [ "import matplotlib.pyplot as plt\nk_list = range(1,101)\naccuracies = []\nfor k in k_list:\n classifier = KNeighborsClassifier(n_neighbors = k)\n classifier.fit(X_train.drop(columns='player_name'), y_train)\n accuracies.append(classifier.score(X_test.drop(columns='player_name'), y_test))\nplt.plot(k_list, accuracies)\nplt.xlabel(\"k\")\nplt.ylabel(\"Validation Accuracy\")\nplt.title(\"Breast Cancer Classifier Accuracy\")\nplt.show()", "_____no_output_____" ] ], [ [ "# K-NN regression 알고리즘", "_____no_output_____" ] ], [ [ "from sklearn.neighbors import KNeighborsRegressor\n", "_____no_output_____" ], [ "regressor = KNeighborsRegressor(n_neighbors = 10, weights = \"distance\")", "_____no_output_____" ], [ "regressor.fit(X_train.drop(columns='player_name'), y_train)\n", "_____no_output_____" ], [ "y_pred = regressor.predict(X_test.drop(columns='player_name'))\ny_pred_train = regressor.predict(X_train.drop(columns='player_name'))", "_____no_output_____" ], [ "result = []\n\nfor i in range(len(y_pred)):\n if y_test[i] < y_pred[i]:\n result.append(\"overpaid\")\n else:\n result.append(\"underpaid\")\n ", "_____no_output_____" ], [ "result", "_____no_output_____" ], [ "len(y_pred)", "_____no_output_____" ], [ "# X_train = X_train.reset_index().drop(columns='index')\nX_test = X_test.reset_index().drop(columns='index')", "_____no_output_____" ], [ "X_train", "_____no_output_____" ], [ "X_test", "_____no_output_____" ], [ "X_test.loc[1]['player_name']", "_____no_output_____" ], [ "underpaid = []\n\nfor i in range(len(y_pred)):\n if result[i] == 'underpaid':\n underpaid.append(X_test.loc[i]['player_name'])", "_____no_output_____" ], [ "len(underpaid)", "_____no_output_____" ], [ "underpaid", "_____no_output_____" ] ], [ [ "# K-NN regression 알고리즘 -> 전체를 다 학습시킴..", "_____no_output_____" ] ], [ [ "from sklearn.neighbors import KNeighborsRegressor\n", "_____no_output_____" ], [ "regressor = KNeighborsRegressor(n_neighbors = 10, weights = \"distance\")", "_____no_output_____" ], [ "regressor.fit(data.drop(columns=['player_name','value']), data.value)\n", "_____no_output_____" ], [ "y_pred = regressor.predict(data.drop(columns=['player_name','value']))", "_____no_output_____" ], [ "result = []\n\nfor i in range(len(y_pred)):\n if data.value[i] < y_pred[i]:\n result.append(\"underpaid\")\n else:\n result.append(\"overpaid\")\n ", "_____no_output_____" ], [ "data.value[1]", "_____no_output_____" ], [ "y_pred[1]", "_____no_output_____" ], [ "result", "_____no_output_____" ], [ "len(y_pred)", "_____no_output_____" ], [ "# X_train = X_train.reset_index().drop(columns='index')\nX_test = X_test.reset_index().drop(columns='index')", "_____no_output_____" ], [ "X_train", "_____no_output_____" ], [ "X_test", "_____no_output_____" ], [ "X_test.loc[1]['player_name']", "_____no_output_____" ], [ "underpaid = []\n\nfor i in range(len(y_pred)):\n if result[i] == 'underpaid':\n underpaid.append(X_test.loc[i]['player_name'])", "_____no_output_____" ], [ "len(underpaid)", "_____no_output_____" ], [ "X_test", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/Skantastico/DS-Unit-2-Applied-Modeling/blob/master/LS_DSPT3_231_Updated_assignment_applied_modeling_1.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ], [ "Lambda School Data Science\n\n*Unit 2, Sprint 3, Module 1*\n\n---\n\n\n# Define ML problems\n\nYou will use your portfolio project dataset for all assignments this sprint.\n\n## Assignment\n\nComplete these tasks for your project, and document your decisions.\n\n- [x] Choose your target. Which column in your tabular dataset will you predict?\n- [x] Is your problem regression or classification?\n- [x] How is your target distributed?\n - Classification: How many classes? Are the classes imbalanced?\n - Regression: Is the target right-skewed? If so, you may want to log transform the target.\n- [x] Choose which observations you will use to train, validate, and test your model.\n - Are some observations outliers? Will you exclude them?\n - Will you do a random split or a time-based split?\n- [x] Choose your evaluation metric(s).\n - Classification: Is your majority class frequency > 50% and < 70% ? If so, you can just use accuracy if you want. Outside that range, accuracy could be misleading. What evaluation metric will you choose, in addition to or instead of accuracy?\n- [ ] Begin to clean and explore your data.\n- [ ] Begin to choose which features, if any, to exclude. Would some features \"leak\" future information?", "_____no_output_____" ] ], [ [ "import pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\n\nurl = 'https://raw.githubusercontent.com/Skantastico/DS-Unit-2-Applied-Modeling/master/data/Anime.csv'\n\ndf = pd.read_csv(url)", "_____no_output_____" ] ], [ [ "# My Dataset\n\nAnime Ratings from the 'iMDB\" of Anime, called myanimelist.net", "_____no_output_____" ] ], [ [ "df.head(7)", "_____no_output_____" ] ], [ [ "## Summary of numeric and non-numeric columns at a glance", "_____no_output_____" ] ], [ [ "df.describe().T", "_____no_output_____" ], [ "df.describe(exclude='number').T", "_____no_output_____" ], [ "col_list = df.columns.values.tolist()\ncol_list", "_____no_output_____" ] ], [ [ "## I was running into trouble during data exploration, there seems to be a space after every column", "_____no_output_____" ] ], [ [ "## I found this piece of code on medium that seems like a catch-all for fixing columns\n\ndf.columns = df.columns.str.strip().str.lower().str.replace(' ', '_').str.replace('(', '').str.replace(')', '')\n\ndf", "_____no_output_____" ], [ "col_list = df.columns.values.tolist()\ncol_list", "_____no_output_____" ], [ "df.columns", "_____no_output_____" ], [ "df.columns.map(lambda x: x.strip())\ndf.columns", "_____no_output_____" ], [ "df.genre.value_counts()", "_____no_output_____" ] ], [ [ "## Ok that seems to have fixed it.\n\nThere seem to be at least around 900 'adult themed' anime which I will probably remove from the dataset, or at least from any public portions just to be safe.\n\nIf it affects the model accuracy at all or is relevant, I will include it for calculations and just make a note.", "_____no_output_____" ], [ "# Choose Your Target", "_____no_output_____" ] ], [ [ "# My Target will be involving the 'score' column\n\ndf.score.value_counts(ascending=False)", "_____no_output_____" ] ], [ [ "## As I will be using the entire spectrum of score, this will be a regression.", "_____no_output_____" ], [ "# How is my target distributed?", "_____no_output_____" ] ], [ [ "# The mean seems to be around 6.3, with only 25% of the dataset above a 7.05\n\ndf.score.describe()", "_____no_output_____" ], [ "df['mean'] = df['score'] >= 6.2845\ndf['mean'].value_counts(normalize=True)", "_____no_output_____" ] ], [ [ "### So there are about 51% anime that are above average (before cleaning)", "_____no_output_____" ], [ "## Which Observations will I use to train?", "_____no_output_____" ], [ "### There's lots of options, but at the very least these look interesting:\n\nNumeric:\n* Episodes\n* Airing\n*Aired\n*Duration\n*Score\n*Popularity\n*Rank\n\nNon-numeric:\n\n* Type\n*Source\n*Producer\n*Genre\n*Studio\n*Rating\n\n", "_____no_output_____" ], [ "## On my old dataset, it was a lot less complete than this one, and I was already thinking of making my own features similar to \"popularity\" and \"# of episodes\".\n\n## I was also going to try and figure out how to scrape and match air dates, but someone other magical person already has! Hooray.", "_____no_output_____" ], [ "## I will continue to clean the dataset on a new notebook", "_____no_output_____" ] ], [ [ "", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## Day 5: Optimal Mind Control\n\nWelcome to Day 6! Now that we can simulate a model network of conductance-based neurons, we discuss the limitations of our approach and attempts to work around these issues.", "_____no_output_____" ], [ "### Memory Management\n\nUsing Python and TensorFlow allowed us to write code that is readable, parallizable and scalable across a variety of computational devices. However, our implementation is very memory intensive. The iterators in TensorFlow do not follow the normal process of memory allocation and garbage collection. Since, TensorFlow is designed to work on diverse hardware like GPUs, TPUs and distributed platforms, memory allocation is done adaptively during the TensorFlow session and not cleared until the Python kernel has stopped execution. The memory used increases linearly with time as the state matrix is computed recursively by the tf.scan function. The maximum memory used by the computational graph is 2 times the total state matrix size at the point when the computation finishes and copies the final data into the memory. The larger the network and longer the simulation, the larger the solution matrix. Each run is limited by the total available memory. For a system with a limited memory of K bytes, The length of a given simulation (L timesteps) of a given network (N differential equations) with 64-bit floating-point precision will follow:\n\n$$2\\times64\\times L\\times N=K$$ \n\nThat is, for any given network, our maximum simulation length is limited. One way to improve our maximum length is to divide the simulation into smaller batches. There will be a small queuing time between batches, which will slow down our code by a small amount but we will be able to simulate longer times. Thus, if we split the simulation into K sequential batches, the maximum memory for the simulation becomes $(1+\\frac{1}{K})$ times the total matrix size. Thus the memory relation becomes: \n\n$$\\Big(1+\\frac{1}{K}\\Big)\\times64\\times L\\times N=K$$ \n\nThis way, we can maximize the length of out simulation that we can run in a single python kernel.\n\nLet us implement this batch system for our 3 neuron feed-forward model.\n\n### Implementing the Model\n\nTo improve the readability of our code we separate the integrator into a independent import module. The integrator code was placed in a file called tf integrator.py. The file must be present in the same directory as the implementation of the model.\n\nNote: If you are using Jupyter Notebook, remember to remove the %matplotlib inline command as it is specific to jupyter.\n\n#### Importing tf_integrator and other requirements\n\nOnce the Integrator is saved in tf_integrator.py in the same directory as the Notebook, we can start importing the essentials including the integrator.", "_____no_output_____" ] ], [ [ "import numpy as np\nimport tf_integrator as tf_int\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n\nimport tensorflow as tf\n## OR ##\n# import tensorflow.compat.v1 as tf\n# tf.disable_v2_behavior()", "_____no_output_____" ] ], [ [ "### Recall the Model\n\nFor implementing a Batch system, we do not need to change how we construct our model only how we execute it.\n\n#### Step 1: Initialize Parameters and Dynamical Equations; Define Input", "_____no_output_____" ] ], [ [ "n_n = 3 # number of simultaneous neurons to simulate\n\nsim_res = 0.01 # Time Resolution of the Simulation\nsim_time = 700 # Length of the Simulation\nt = np.arange(0,sim_time,sim_res) \n\n# Acetylcholine\n\nach_mat = np.zeros((n_n,n_n)) # Ach Synapse Connectivity Matrix\nach_mat[1,0]=1\n\n## PARAMETERS FOR ACETYLCHLOLINE SYNAPSES ##\n\nn_ach = int(np.sum(ach_mat)) # Number of Acetylcholine (Ach) Synapses \nalp_ach = [10.0]*n_ach # Alpha for Ach Synapse\nbet_ach = [0.2]*n_ach # Beta for Ach Synapse\nt_max = 0.3 # Maximum Time for Synapse\nt_delay = 0 # Axonal Transmission Delay\nA = [0.5]*n_n # Synaptic Response Strength\ng_ach = [0.35]*n_n # Ach Conductance\nE_ach = [0.0]*n_n # Ach Potential\n\n# GABAa\n\ngaba_mat = np.zeros((n_n,n_n)) # GABAa Synapse Connectivity Matrix\ngaba_mat[2,1] = 1\n\n## PARAMETERS FOR GABAa SYNAPSES ##\n\nn_gaba = int(np.sum(gaba_mat)) # Number of GABAa Synapses\nalp_gaba = [10.0]*n_gaba # Alpha for GABAa Synapse\nbet_gaba = [0.16]*n_gaba # Beta for GABAa Synapse\nV0 = [-20.0]*n_n # Decay Potential\nsigma = [1.5]*n_n # Decay Time Constant\ng_gaba = [0.8]*n_n # fGABA Conductance\nE_gaba = [-70.0]*n_n # fGABA Potential\n\n## Storing Firing Thresholds ##\n\nF_b = [0.0]*n_n # Fire threshold\n\ndef I_inj_t(t):\n return tf.constant(current_input.T,dtype=tf.float64)[tf.to_int32(t/sim_res)] # Turn indices to integer and extract from matrix\n\n## Acetylcholine Synaptic Current ##\n\ndef I_ach(o,V):\n o_ = tf.Variable([0.0]*n_n**2,dtype=tf.float64)\n ind = tf.boolean_mask(tf.range(n_n**2),ach_mat.reshape(-1) == 1)\n o_ = tf.scatter_update(o_,ind,o)\n o_ = tf.transpose(tf.reshape(o_,(n_n,n_n)))\n return tf.reduce_sum(tf.transpose((o_*(V-E_ach))*g_ach),1)\n\n## GABAa Synaptic Current ##\n\ndef I_gaba(o,V):\n o_ = tf.Variable([0.0]*n_n**2,dtype=tf.float64)\n ind = tf.boolean_mask(tf.range(n_n**2),gaba_mat.reshape(-1) == 1)\n o_ = tf.scatter_update(o_,ind,o)\n o_ = tf.transpose(tf.reshape(o_,(n_n,n_n)))\n return tf.reduce_sum(tf.transpose((o_*(V-E_gaba))*g_gaba),1)\n\n## Other Currents ##\n\ndef I_K(V, n):\n return g_K * n**4 * (V - E_K)\n\ndef I_Na(V, m, h):\n return g_Na * m**3 * h * (V - E_Na)\n\ndef I_L(V):\n return g_L * (V - E_L)\n\ndef dXdt(X, t):\n V = X[:1*n_n] # First n_n values are Membrane Voltage\n m = X[1*n_n:2*n_n] # Next n_n values are Sodium Activation Gating Variables\n h = X[2*n_n:3*n_n] # Next n_n values are Sodium Inactivation Gating Variables\n n = X[3*n_n:4*n_n] # Next n_n values are Potassium Gating Variables\n o_ach = X[4*n_n : 4*n_n + n_ach] # Next n_ach values are Acetylcholine Synapse Open Fractions\n o_gaba = X[4*n_n + n_ach : 4*n_n + n_ach + n_gaba] # Next n_ach values are GABAa Synapse Open Fractions\n fire_t = X[-n_n:] # Last n_n values are the last fire times as updated by the modified integrator\n \n dVdt = (I_inj_t(t) - I_Na(V, m, h) - I_K(V, n) - I_L(V) - I_ach(o_ach,V) - I_gaba(o_gaba,V)) / C_m \n \n ## Updation for gating variables ##\n \n m0,tm,h0,th = Na_prop(V)\n n0,tn = K_prop(V)\n\n dmdt = - (1.0/tm)*(m-m0)\n dhdt = - (1.0/th)*(h-h0)\n dndt = - (1.0/tn)*(n-n0)\n \n ## Updation for o_ach ##\n \n A_ = tf.constant(A,dtype=tf.float64)\n Z_ = tf.zeros(tf.shape(A_),dtype=tf.float64)\n \n T_ach = tf.where(tf.logical_and(tf.greater(t,fire_t+t_delay),tf.less(t,fire_t+t_max+t_delay)),A_,Z_) \n T_ach = tf.multiply(tf.constant(ach_mat,dtype=tf.float64),T_ach)\n T_ach = tf.boolean_mask(tf.reshape(T_ach,(-1,)),ach_mat.reshape(-1) == 1)\n \n do_achdt = alp_ach*(1.0-o_ach)*T_ach - bet_ach*o_ach\n \n ## Updation for o_gaba ##\n \n T_gaba = 1.0/(1.0+tf.exp(-(V-V0)/sigma))\n T_gaba = tf.multiply(tf.constant(gaba_mat,dtype=tf.float64),T_gaba)\n T_gaba = tf.boolean_mask(tf.reshape(T_gaba,(-1,)),gaba_mat.reshape(-1) == 1)\n \n do_gabadt = alp_gaba*(1.0-o_gaba)*T_gaba - bet_gaba*o_gaba\n \n ## Updation for fire times ##\n \n dfdt = tf.zeros(tf.shape(fire_t),dtype=fire_t.dtype) # zero change in fire_t\n \n\n out = tf.concat([dVdt,dmdt,dhdt,dndt,do_achdt,do_gabadt,dfdt],0)\n return out\n\ndef K_prop(V):\n T = 22\n phi = 3.0**((T-36.0)/10)\n V_ = V-(-50)\n \n alpha_n = 0.02*(15.0 - V_)/(tf.exp((15.0 - V_)/5.0) - 1.0)\n beta_n = 0.5*tf.exp((10.0 - V_)/40.0)\n \n t_n = 1.0/((alpha_n+beta_n)*phi)\n n_0 = alpha_n/(alpha_n+beta_n)\n \n return n_0, t_n\n\n\ndef Na_prop(V):\n T = 22\n phi = 3.0**((T-36)/10)\n V_ = V-(-50)\n \n alpha_m = 0.32*(13.0 - V_)/(tf.exp((13.0 - V_)/4.0) - 1.0)\n beta_m = 0.28*(V_ - 40.0)/(tf.exp((V_ - 40.0)/5.0) - 1.0)\n \n alpha_h = 0.128*tf.exp((17.0 - V_)/18.0)\n beta_h = 4.0/(tf.exp((40.0 - V_)/5.0) + 1.0)\n \n t_m = 1.0/((alpha_m+beta_m)*phi)\n t_h = 1.0/((alpha_h+beta_h)*phi)\n \n m_0 = alpha_m/(alpha_m+beta_m)\n h_0 = alpha_h/(alpha_h+beta_h)\n \n return m_0, t_m, h_0, t_h\n\n\n# Initializing the Parameters\n\nC_m = [1.0]*n_n\ng_K = [10.0]*n_n\nE_K = [-95.0]*n_n\n\ng_Na = [100]*n_n\nE_Na = [50]*n_n \n\ng_L = [0.15]*n_n\nE_L = [-55.0]*n_n\n\n# Creating the Current Input\ncurrent_input= np.zeros((n_n,t.shape[0]))\ncurrent_input[0,int(100/sim_res):int(200/sim_res)] = 2.5\ncurrent_input[0,int(300/sim_res):int(400/sim_res)] = 5.0\ncurrent_input[0,int(500/sim_res):int(600/sim_res)] = 7.5", "_____no_output_____" ] ], [ [ "#### Step 2: Define the Initial Condition of the Network and Add some Noise to the initial conditions", "_____no_output_____" ] ], [ [ "# Initializing the State Vector and adding 1% noise\nstate_vector = [-71]*n_n+[0,0,0]*n_n+[0]*n_ach+[0]*n_gaba+[-9999999]*n_n\nstate_vector = np.array(state_vector)\nstate_vector = state_vector + 0.01*state_vector*np.random.normal(size=state_vector.shape)", "_____no_output_____" ] ], [ [ "#### Step 3: Splitting Time Series into independent batches and Run Each Batch Sequentially\n\nSince we will be dividing the computation into batches, we have to split the time array such that for each new call, the final state vector of the last batch will be the initial condition for the current batch. The function $np.array\\_split()$ splits the array into non-overlapping vectors. Therefore, we append the last time of the previous batch to the beginning of the current time array batch.", "_____no_output_____" ] ], [ [ "# Define the Number of Batches\nn_batch = 2\n\n# Split t array into batches using numpy\nt_batch = np.array_split(t,n_batch)\n\n# Iterate over the batches of time array\nfor n,i in enumerate(t_batch):\n \n # Inform start of Batch Computation\n print(\"Batch\",(n+1),\"Running...\",end=\"\")\n \n # In np.array_split(), the split edges are present in only one array and since \n # our initial vector to successive calls is corresposnding to the last output\n # our first element in the later time array should be the last element of the \n # previous output series, Thus, we append the last time to the beginning of \n # the current time array batch.\n if n>0:\n i = np.append(i[0]-sim_res,i)\n \n # Set state_vector as the initial condition\n init_state = tf.constant(state_vector, dtype=tf.float64)\n # Create the Integrator computation graph over the current batch of t array\n tensor_state = tf_int.odeint(dXdt, init_state, i, n_n, F_b)\n \n # Initialize variables and run session\n with tf.Session() as sess:\n tf.global_variables_initializer().run()\n state = sess.run(tensor_state)\n sess.close()\n \n # Reset state_vector as the last element of output\n state_vector = state[-1,:]\n \n # Save the output of the simulation to a binary file\n np.save(\"part_\"+str(n+1),state)\n\n # Clear output\n state=None\n \n print(\"Finished\")", "Batch 1 Running...Finished\nBatch 2 Running...Finished\n" ] ], [ [ "#### Putting the Output Together\n\nThe output from our batch implementation is a set of binary files that store parts of our total simulation. To get the overall output we have to stitch them back together.", "_____no_output_____" ] ], [ [ "overall_state = []\n\n# Iterate over the generated output files\nfor n,i in enumerate([\"part_\"+str(n+1)+\".npy\" for n in range(n_batch)]):\n \n # Since the first element in the series was the last output, we remove them\n if n>0:\n overall_state.append(np.load(i)[1:,:])\n else:\n overall_state.append(np.load(i))\n\n# Concatenate all the matrix to get a single state matrix\noverall_state = np.concatenate(overall_state)", "_____no_output_____" ] ], [ [ "#### Visualizing the Overall Data\nFinally, we plot the voltage traces of the 3 neurons as a Voltage vs Time heatmap.", "_____no_output_____" ] ], [ [ "plt.figure(figsize=(12,6))\n \nsns.heatmap(overall_state[::100,:3].T,xticklabels=100,yticklabels=5,cmap='RdBu_r')\n\nplt.xlabel(\"Time (in ms)\")\nplt.ylabel(\"Neuron Number\")\nplt.title(\"Voltage vs Time Heatmap for Projection Neurons (PNs)\")\n\nplt.tight_layout()\nplt.show()", "_____no_output_____" ] ], [ [ "By this method, we have maximized the usage of our available memory but we can go further and develop a method to allow indefinitely long simulation. The issue behind this entire algorithm is that the memory is not cleared until the python kernel finishes. One way to overcome this is to save the parameters of the model (such as connectivity matrix) and the state vector in a file, and start a new python kernel from a python script to compute successive batches. This way after each large batch, the memory gets cleaned. By combining the previous batch implementation and this system, we can maximize our computability.\n\n### Implementing a Runner and a Caller\n\nFirstly, we have to create an implementation of the model that takes in previous input as current parameters. Thus, we create a file, which we call \"run.py\" that takes an argument ie. the current batch number. The implementation for \"run.py\" is mostly same as the above model but there is a small difference.\n\nWhen the batch number is 0, we initialize all variable parameters and save them, but otherwise we use the saved values. The parameters we save include: Acetylcholine Matrix, GABAa Matrix and Final/Initial State Vector. It will also save the files with both batch number and sub-batch number listed.\n\nThe time series will be created and split initially by the caller, which we call \"call.py\", and stored in a file. Each execution of the Runner will extract its relevant time series and compute on it.", "_____no_output_____" ], [ "#### Implementing the Caller code\n\nThe caller will create the time series, split it and use python subprocess module to call \"run.py\" with appropriate arguments. The code for \"call.py\" is given below.", "_____no_output_____" ] ], [ [ "from subprocess import call\nimport numpy as np\n\ntotal_time = 700\nn_splits = 2\ntime = np.split(np.arange(0,total_time,0.01),n_splits)\n\n# Append the last time point to the beginning of the next batch\nfor n,i in enumerate(time):\n if n>0:\n time[n] = np.append(i[0]-0.01,i)\n\nnp.save(\"time\",time)\n\n# call successive batches with a new python subprocess and pass the batch number\nfor i in range(n_splits):\n call(['python','run.py',str(i)])\n\nprint(\"Simulation Completed.\")", "_____no_output_____" ] ], [ [ "#### Implementing the Runner code\n\n\"run.py\" is essentially identical to the batch-implemented model we developed above with the changes described below:", "_____no_output_____" ] ], [ [ "# Additional Imports #\n\nimport sys\n\n# Duration of Simulation #\n\n# t = np.arange(0,sim_time,sim_res) \nt = np.load(\"time.npy\")[int(sys.argv[1])] # get first argument to run.py\n\n# Connectivity Matrix Definitions #\n\nif sys.argv[1] == '0':\n ach_mat = np.zeros((n_n,n_n)) # Ach Synapse Connectivity Matrix\n ach_mat[1,0]=1 # If connectivity is random, once initialized it will be the same.\n np.save(\"ach_mat\",ach_mat)\nelse:\n ach_mat = np.load(\"ach_mat.npy\")\n \nif sys.argv[1] == '0':\n gaba_mat = np.zeros((n_n,n_n)) # GABAa Synapse Connectivity Matrix\n gaba_mat[2,1] = 1 # If connectivity is random, once initialized it will be the same.\n np.save(\"gaba_mat\",gaba_mat)\nelse:\n gaba_mat = np.load(\"gaba_mat.npy\")\n\n# Current Input Definition #\n \nif sys.argv[1] == '0':\n current_input= np.zeros((n_n,int(sim_time/sim_res)))\n current_input[0,int(100/sim_res):int(200/sim_res)] = 2.5\n current_input[0,int(300/sim_res):int(400/sim_res)] = 5.0\n current_input[0,int(500/sim_res):int(600/sim_res)] = 7.5\n np.save(\"current_input\",current_input)\nelse:\n current_input = np.load(\"current_input.npy\")\n \n# State Vector Definition #\n\nif sys.argv[1] == '0':\n sstate_vector = [-71]*n_n+[0,0,0]*n_n+[0]*n_ach+[0]*n_gaba+[-9999999]*n_n\n state_vector = np.array(state_vector)\n state_vector = state_vector + 0.01*state_vector*np.random.normal(size=state_vector.shape)\n np.save(\"state_vector\",state_vector)\nelse:\n state_vector = np.load(\"state_vector.npy\")\n\n# Saving of Output #\n\n# np.save(\"part_\"+str(n+1),state)\nnp.save(\"batch\"+str(int(sys.argv[1])+1)+\"_part_\"+str(n+1),state)", "_____no_output_____" ] ], [ [ "#### Combining all Data\n\nJust like we merged all the batches, we merge all the sub-batches and batches.", "_____no_output_____" ] ], [ [ "overall_state = []\n\n# Iterate over the generated output files\nfor n,i in enumerate([\"batch\"+str(x+1) for x in range(n_splits)]):\n for m,j in enumerate([\"_part_\"+str(x+1)+\".npy\" for x in range(n_batch)]):\n \n # Since the first element in the series was the last output, we remove them\n if n>0 and m>0:\n overall_state.append(np.load(i+j)[1:,:])\n else:\n overall_state.append(np.load(i+j))\n\n# Concatenate all the matrix to get a single state matrix\noverall_state = np.concatenate(overall_state)", "_____no_output_____" ], [ "plt.figure(figsize=(12,6))\n \nsns.heatmap(overall_state[::100,:3].T,xticklabels=100,yticklabels=5,cmap='RdBu_r')\n\nplt.xlabel(\"Time (in ms)\")\nplt.ylabel(\"Neuron Number\")\nplt.title(\"Voltage vs Time Heatmap for Projection Neurons (PNs)\")\n\nplt.tight_layout()\nplt.show()", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "%matplotlib inline", "_____no_output_____" ] ], [ [ "Using $L_0$ regularization in predicting genetic risk\n====================================\n\nThe main aim of this document is to outline the code and theory of using the $L_0$ norm in a regularized regression with the objective to predict disease risk from genetic data.\n\nThis document contains my thought process and understanding of the implementation of $L_0$.\nI will aim to integrate the code into existing tools so it can be easily used.\nIdeally I will also compare the performance again $L_1$ as well as $L_2$ norm (aka Lasso and Ridge regression) in predicting simulated data.\n\n## Literature used\n\nI made use of the following papers:\n\n- Learning Sparse Neural Networks through $L_0$ Regularization [link](http://arxiv.org/abs/1712.01312)\n- The Variational Garrote [link](http://link.springer.com/10.1007/s10994-013-5427-7)\n\n# Introduction\n\nLet $\\mathcal{D}$ the dataset consisting of $N$ input-output pairs $\\{(x_1, y_1), \\ldots, (x_N, y_N)\\}$ and consider the regularized minimization procedure with $L_p$ regularization of the parameters $\\omega$\n\n$$\n\\mathcal{R}(\\theta) = \\frac{1}{N} ( \\sum^N_{i=1} \\mathcal{L}(h(x_i; \\theta), y_i)) + \\lambda ||\\theta||_p\n$$\n\nwith $\\theta^* = arg \\min_{\\theta}\\{\\mathcal{R}(\\theta)\\}$.\n\nWhile $L_1$ and $L_2$ are well defined the definition of $L_0$ varies.\nHence I will define it as:\n$$\n||\\theta||_0 = \\sum^{|\\theta|}_{j=1} I[\\theta_j \\neq 0]\n$$\n\nHere it is important that, in contrast to $L_1$ and $L_2$, the $L_0$ norm does not shrink the parameters value but sets it wither to 0 or not 0.\nThis effect might be desirable in genetic data due to high correlations among SNPs.", "_____no_output_____" ] ], [ [ "import numpy as np\nimport matplotlib.pyplot as plt\nplt.style.use('fivethirtyeight')", "_____no_output_____" ], [ "def l0(x):\n return(np.sum(x!=0))\n\ndef l1(x):\n return(np.sum(np.abs(x)))\n\ndef l2(x):\n return(np.sum(np.power(x, 2)))", "_____no_output_____" ], [ "x = np.linspace(-2, 2, 50)\nx = np.append(x, 0)\nx = np.sort(x)\n\nfig, (axs0, axs1, axs2) = plt.subplots(1, 3, sharex='all', sharey='all')\nfig.set_size_inches((10, 3))\n\naxs0.set_title('L0 norm')\naxs1.set_title('L1 norm')\naxs2.set_title('L2 norm')\n\n[k.set_xlabel(r'$\\theta$') for k in [axs0, axs1, axs2]]\n[k.set_ylabel('Penalty') for k in [axs0, axs1, axs2]]\n\naxs0.plot(x, [l0(k) for k in x])\naxs1.plot(x, [l1(k) for k in x])\naxs2.plot(x, [l2(k) for k in x])", "_____no_output_____" ] ], [ [ "The plot above demonstrates nicely the penalty for different norms.\nAs one can see both $p=1$ and $p=2$ allow shrinkage for large values of $\\theta$, while $p=0$ the penalty is constant.", "_____no_output_____" ], [ "## Minimizing $L_0$ norm for parametric models\n\nOptimization under the $L_0$ penalty is computational difficult due to the non-differentiability as well as the comninatorial nature of $2^{|\\theta|}$ (the number of possible states of the $L_0$ penalty function).\n\nHence @Author relaxed the discrete nature of $L_0$ to allow for efficient computations.\nHere I will outline their approach first in order to understand it better.\n\nThe first step is to reformulate the $L_0$ norm under the parameters $\\theta$.\nHence let,\n$$\n\\begin{matrix}\n\\theta_j = \\tilde{\\theta_j}z_j, & z_j \\in \\{0, 1\\}, & \\tilde{\\theta}_j \\neq 0, & ||\\theta||_0 = \\sum^{|\\theta|}_{j=1} z_j\n\\end{matrix}\n$$\n\nTherefore $z_j$ can be considered as binary gates, and the $L_0$ norm can be seen as the amount of gates being turned on.\nThen we can reformulate the minimization from Eq. 1 by letting $q(z_j|\\pi_j) = Bern(\\pi_j)$\n$$\n\\mathcal{R}(\\tilde{\\theta}, \\pi) = \\mathbb{E}_{q(z|\\pi)} [\\frac{1}{N} ( \\sum^N_{i=1} \\mathcal{L}(h(x_i; \\tilde{\\theta} \\otimes z), y_i)] + \\lambda \\sum^{|\\theta|}_{j=1} \\pi_j \\\\\n\\quad \\textrm{with} \\quad \n\\theta^*, \\pi^* = arg \\min_{\\tilde{\\theta}, \\pi} \\{\\mathcal{R}(\\tilde{\\theta}, \\pi)\\}\n$$\n\nHere we encounter the problem of non-differentiability of the left hand side of the previous equation size z is discrete (the penalty term is straight forward to minimize).\nIt seems there are methods to optimize this functions but they are supposed to suffer from high variance and biased gradients.\nTherefore @Author used an alternative to smooth the objective function for efficient gradient based optimization.\n\nLet $s$ be a continuous random variable with distribution $q(s)$ that has the parameter $\\phi$ we can then let $z$ be given a hard sigmoid function:\n$$\ns \\sim q(s|\\phi) \\\\\nz = min(1, max(0, s))\n$$\nThis allows the gate to still be $0$ as well as the computation of the probability of non-zero parameters (number of active gates) with a simple cdf ($Q(\\cdot)$ of s:\n$$\nq(z\\neq 0 | \\phi) = 1 - Q(s \\leq 0 | \\phi)\n$$\nThen we can reformulate the previous equation as\n$$\n\\mathcal{R}(\\tilde{\\theta}, \\phi) = \\mathbb{E}_{q(s|\\phi)} [\\frac{1}{N} ( \\sum^N_{i=1} \\mathcal{L}(h(x_i; \\tilde{\\theta} \\otimes g(s)), y_i)] + \\lambda \\sum^{|\\theta|}_{j=1} (1 - Q(s_j \\leq 0 | \\phi_j)) \\\\\n\\quad \\textrm{with} \\quad \n\\theta^*, \\phi^* = arg \\min_{\\tilde{\\theta}, \\phi} \\{\\mathcal{R}(\\tilde{\\theta}, \\phi)\\},\n\\quad \ng(\\cdot) = min(1, max(0, \\cdot))\n$$\nHowever, it is still necessary to define the distribution for $s$.\nHere it is suggested to use a hard concrete distribution.", "_____no_output_____" ], [ "### The hard concrete distribution\n\nLets assume that $s$ is a random variable distributed in the (0, 1) interval with probability density $q_s(s|\\phi)$ and cdf $Q_s(s|\\phi)$.\nThe parameters of the distribution are $\\phi = (\\log \\alpha, \\beta)$, where $\\log\\alpha$ is the mean and $\\beta$ the temperature (???).\nWe can then stretch this distribution to the $(\\gamma, \\delta$) interval with $\\gamma < 0$ and $\\delta > 1$:\n$$\nu\\sim U(0, 1), \n\\quad\ns = Sigmoid((\\log u - \\log(1 - u) + \\log\\alpha) / \\beta), \\quad\n\\bar{s} = s(\\delta - \\gamma) + \\gamma, \\quad\nz = \\min(1, \\max(0, \\bar{s})\n$$\n\nSince these formulas are a bit difficult to visualize I have plotted them below to gain a better understanding of that they are doing and how these functions behave.", "_____no_output_____" ] ], [ [ "def hard_sigmoid(x):\n return np.min([1, np.max([0, x])])\n\ndef sigmoid(x):\n return 1 / (1 + np.exp(-x))\n\ndef hard_concrete_dist(loc, temp, gamma, zeta):\n u = np.random.random()\n s = sigmoid((np.log(u) - np.log(1 - u) + np.log(loc)) / temp)\n shat = s*(zeta - gamma) + gamma\n return hard_sigmoid(shat)\n\ndef plot_hard_concreate(loc, temp, gamma, zeta, num=10000, bins=100, **kwargs):\n plt.hist([hard_concrete_dist(loc, temp, gamma, zeta) for _ in range(num)], bins=bins, density=True, **kwargs)", "_____no_output_____" ], [ "x = np.arange(-6, 6, 0.1)\ny1 = sigmoid(x)\ny2 = [hard_sigmoid(k) for k in x]", "_____no_output_____" ], [ "fig, axs = plt.subplots(1, 2)\nfig.set_size_inches((10, 3))\naxs[0].set_title('Comparision of Sigmoids')\naxs[1].set_title('Hard Concreate Distribution')\n\naxs[0].plot(x, y1, label='Sigmoid')\naxs[0].plot(x, y2, label='Hard-Sigmoid')\nhandles, labels = axs[0].get_legend_handles_labels()\naxs[0].legend(handles, labels)\nvalues = [hard_concrete_dist(loc=1, temp=0.5, gamma=-0.1, zeta=1.1) for _ in range(1000)]\nn, bins, patches = axs[1].hist(values, bins=100, density=True)\n#fig.show()", "_____no_output_____" ], [ "fig, axs = plt.subplots()\nfig.set_size_inches((10, 8))\naxs.set_title('Comparision of Sigmoids')\n\naxs.plot(x, y1, label='Sigmoid')\naxs.plot(x, y2, label='Hard-Sigmoid')\nhandles, labels = axs.get_legend_handles_labels()\naxs.legend(handles, labels)", "_____no_output_____" ], [ "fig, axs = plt.subplots()\nfig.set_size_inches((10, 5))\naxs.set_title('Hard Concreate Distribution')\n\nvalues = [hard_concrete_dist(loc=1, temp=0.5, gamma=-0.1, zeta=1.1) for _ in range(5000)]\nn, bins, patches = axs.hist(values, bins=100, density=True)", "_____no_output_____" ] ], [ [ "# Implementation of the $L_0$ norm\n\nThe next step is to implement the theory into practice.\nI will therefore make use of Google's tensorflow to implement the $L_0$ norm.\nHere its good to know that this has been implemented before in PyTorch.\nI will compare my and their implementation to assure I have done it correctly.\nThe reason why I am not directly using PyTorch is that I don't want to learn yet again another library.\nFurthermore, implementing $L_0$ into tensorflow gives me hopefully a better understanding of the library itself as well as the $L_0$ norm.\nIn addition, tensorflow can compute the derivatives for me, so I don't have to take care of this.\n\nThe data I am going to use was simulated by Tim.\nThis includes non-linear effects, binary and continous phenotypes.\nI will first compute only penalized linear models (that is $L_0$ to $L_2$ norm) with tensorflow and validate the results (of $L_1$ and $L_2$ with already implemented libraries (sklearn).\nThe main objective is to improve prediction accuracy.", "_____no_output_____" ] ], [ [ "import tensorflow as tf\nfrom sklearn.model_selection import train_test_split\nfrom pyplink import PyPlink\nimport sys\nimport os", "_____no_output_____" ], [ "DATAFOLDER = os.path.realpath(filename='../data')\nPLINKDATA = '1kgb'\nFILEPATH = os.path.join(DATAFOLDER, PLINKDATA)\n\ndef count_lines(filepath, header=False):\n \"\"\"Count the number of rows in a file\"\"\"\n with open(filepath, 'r') as f:\n count = sum(1 for line in f)\n if header:\n count -= 1\n return count\n\nprint('Number of subjects:', count_lines(os.path.join(FILEPATH, '.fam'), True))\nprint('Number of SNPs:', count_lines(os.path.join(FILEPATH, '.bim'), False))", "_____no_output_____" ], [ "def get_matrix(pfile, max_block):\n \"\"\"Extract a genotype matrix from plink file.\"\"\"\n with PyPlink(pfile) as bed:\n bim = bed.get_bim()\n fam = bed.get_fam()\n n = fam.shape[0]\n p = bim.shape[0]\n assert(max_block <= p)\n genotypemat = np.zeros((n, max_block), dtype=np.int64)\n u = 0\n for loci_name, genotypes in bed:\n genotypemat[:, u] = np.array(genotypes)\n u += 1\n if u >= max_block:\n break\n return genotypemat", "_____no_output_____" ], [ "# define size of training, validation and testing data\nsample_fractions = (0.8, 0.15, 0.05)\nassert sum(sample_fractions) == 1.0\ntraining_size, validation_size, testing_size = sample_fractions\ndata_training, data_testing, label_training, label_testing = train_test_split(pheno, genotypemat, train_size=training_size)\ndata_validataion, data_testing, label_validation, label_testing = train_test_split(data_testing, label_testing, train_size=2/3)\nprint(\"Samples in training:\", len(label_training))\nprint(\"Samples in validation:\", len(label_validation))\nprint(\"Samples in testing:\", len(label_testing))", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "### Regular Expressions\n\nRegular expressions are `text matching patterns` described with a formal syntax. You'll often hear regular expressions referred to as 'regex' or 'regexp' in conversation. Regular expressions can include a variety of rules, for finding repetition, to text-matching, and much more. As you advance in Python you'll see that a lot of your parsing problems can be solved with regular expressions (they're also a common interview question!).\n", "_____no_output_____" ], [ "## Searching for Patterns in Text\n\nOne of the most common uses for the re module is for finding patterns in text. Let's do a quick example of using the search method in the re module to find some text:", "_____no_output_____" ] ], [ [ "import re", "_____no_output_____" ], [ "# List of patterns to search for\npatterns = [ 'term1', 'term2' ]\n\n# Text to parse\ntext = 'This is a string with term1, but it does not have the other term.'\n", "_____no_output_____" ], [ "for p in patterns:\n print ('Searching for \"%s\" in Sentence: \\n\"%s\"' % (p, text))\n \n #Check for match\n if re.search(p, text):\n print ('Match was found. \\n')\n else:\n print ('No Match was found. \\n')", "Searching for \"term1\" in Sentence: \n\"This is a string with term1, but it does not have the other term.\"\nMatch was found. \n\nSearching for \"term2\" in Sentence: \n\"This is a string with term1, but it does not have the other term.\"\nNo Match was found. \n\n" ] ], [ [ "Now we've seen that re.search() will take the pattern, scan the text, and then returns a **Match** object. If no pattern is found, a **None** is returned. To give a clearer picture of this match object, check out the cell below:", "_____no_output_____" ] ], [ [ "# List of patterns to search for\npattern = 'term1'\n\n# Text to parse\ntext = 'This is a string with term1, but it does not have the other term.'\n\nmatch = re.search(pattern, text)\n\ntype(match)", "_____no_output_____" ], [ "match", "_____no_output_____" ] ], [ [ "This **Match** object returned by the search() method is more than just a Boolean or None, it contains information about the match, including the original input string, the regular expression that was used, and the location of the match. Let's see the methods we can use on the match object:", "_____no_output_____" ] ], [ [ "# Show start of match\nmatch.start()", "_____no_output_____" ], [ "# Show end\nmatch.end()", "_____no_output_____" ], [ "s = \"abassabacdReddyceaabadjfvababaReddy\"\nr = re.compile(\"Reddy\")\nr", "_____no_output_____" ], [ "l = re.findall(r,s)\nprint(l)", "['Reddy', 'Reddy']\n" ], [ "import re\ns = \"abcdefg1234\"\nr = re.compile(\"^[a-z][0-9]$\")\nl = re.findall(r,s)\nprint(l)", "[]\n" ], [ "s = \"ABCDE1234a\"\nr = re.compile(r\"^[A-Z]{5}[0-9]{4}[a-z]$\")\nl = re.findall(r,s)\nprint(l)", "['ABCDE1234a']\n" ], [ "s = \"+917123456789\"\ns1 = \"07123456789\"\ns2 = \"7123456789\"\nr = re.compile(r\"[6-9][0-9]{9}\")\nl = re.findall(r,s)\nprint(l)", "['9171234567']\n" ], [ "l = re.findall(r,s1)\nprint(l)", "['7123456789']\n" ], [ "l = re.findall(r,s2)\nprint(l)", "['7123456789']\n" ], [ "s = \"+917234567891\"\ns1 = \"07123456789\"\ns2 = \"7123456789\"\nr = re.compile(r\"^(\\+91)?[0]?([6-9][0-9]{9})$\")\nm = re.search(r,s1)\nif m:\n print(m.group())\nelse:\n print(\"Invalid string\")", "07123456789\n" ], [ "for _ in range(int(input(\"No of Test Cases:\"))):\n line = input(\"Mobile Number\")\n if re.match(r\"^[789]{1}\\d{9}$\", line):\n print(\"YES\")\n else:\n print(\"NO\")", "No of Test Cases: 2\nMobile Number 9666302750\n" ], [ "#Named groups\ns = \"12-02-2017\" # DD-MM-YYYY\n# dd-mm-yyyy\nr = re.compile(r\"^(?P<day>\\d{2})-(?P<month>[0-9]{2})-(?P<year>[0-9]{4})\")\nm = re.search(r,s)\nif m:\n print(m.group('year'))\n print(m.group('month'))\n print(m.group('day'))", "_____no_output_____" ] ], [ [ "## Split with regular expressions\n\nLet's see how we can split with the re syntax. This should look similar to how you used the split() method with strings.", "_____no_output_____" ] ], [ [ "# Term to split on\nsplit_term = '@'\n\nphrase = 'What is the domain name of someone with the email: hello@gmail.com'\n\n# Split the phrase\nre.split(split_term,phrase)", "_____no_output_____" ] ], [ [ "Note how re.split() returns a list with the term to spit on removed and the terms in the list are a split up version of the string. Create a couple of more examples for yourself to make sure you understand!\n\n## Finding all instances of a pattern\n\nYou can use re.findall() to find all the instances of a pattern in a string. For example:", "_____no_output_____" ] ], [ [ "# Returns a list of all matches\nre.findall('is','test phrase match is in middle')", "_____no_output_____" ], [ "a = \" a list with the term to spit on removed and the terms in the list are a split up version of the string. Create a couple of more examples for yourself to make sure you understand!\"\ncopy = re.findall(\"to\",a)\ncopy", "_____no_output_____" ], [ "len(copy)", "_____no_output_____" ] ], [ [ "## Pattern re Syntax\n\nThis will be the bulk of this lecture on using re with Python. Regular expressions supports a huge variety of patterns the just simply finding where a single string occurred. \n\nWe can use *metacharacters* along with re to find specific types of patterns. \n\nSince we will be testing multiple re syntax forms, let's create a function that will print out results given a list of various regular expressions and a phrase to parse:", "_____no_output_____" ] ], [ [ "def multi_re_find(patterns,phrase):\n '''\n Takes in a list of regex patterns\n Prints a list of all matches\n '''\n for pattern in patterns:\n print ('Searching the phrase using the re check: %r' %pattern)\n print (re.findall(pattern,phrase))", "_____no_output_____" ] ], [ [ "### Repetition Syntax\n\nThere are five ways to express repetition in a pattern:\n\n 1.) A pattern followed by the meta-character * is repeated zero or more times. \n 2.) Replace the * with + and the pattern must appear at least once. \n 3.) Using ? means the pattern appears zero or one time. \n 4.) For a specific number of occurrences, use {m} after the pattern, where m is replaced with the number of times the pattern should repeat. \n 5.) Use {m,n} where m is the minimum number of repetitions and n is the maximum. Leaving out n ({m,}) means the value appears at least m times, with no maximum.\n \nNow we will see an example of each of these using our multi_re_find function:", "_____no_output_____" ] ], [ [ "test_phrase = 'sdsd..sssddd...sdddsddd...dsds...dsssss...sdddd'\n\ntest_patterns = [ 'sd*', # s followed by zero or more d's\n 'sd+', # s followed by one or more d's\n 'sd?', # s followed by zero or one d's\n 'sd{3}', # s followed by three d's\n 'sd{2,3}', # s followed by two to three d's\n ]\n\nmulti_re_find(test_patterns,test_phrase)", "Searching the phrase using the re check: 'sd*'\n['sd', 'sd', 's', 's', 'sddd', 'sddd', 'sddd', 'sd', 's', 's', 's', 's', 's', 's', 'sdddd']\nSearching the phrase using the re check: 'sd+'\n['sd', 'sd', 'sddd', 'sddd', 'sddd', 'sd', 'sdddd']\nSearching the phrase using the re check: 'sd?'\n['sd', 'sd', 's', 's', 'sd', 'sd', 'sd', 'sd', 's', 's', 's', 's', 's', 's', 'sd']\nSearching the phrase using the re check: 'sd{3}'\n['sddd', 'sddd', 'sddd', 'sddd']\nSearching the phrase using the re check: 'sd{2,3}'\n['sddd', 'sddd', 'sddd', 'sddd']\n" ] ], [ [ "## Character Sets\n\nCharacter sets are used when you wish to match any one of a group of characters at a point in the input. Brackets are used to construct character set inputs. For example: the input [ab] searches for occurrences of either a or b.\nLet's see some examples:", "_____no_output_____" ] ], [ [ "test_phrase = 'sdsd..sssddd...sdddsddd...dsds...dsssss...sdddd'\n\ntest_patterns = [ '[sd]', # either s or d\n 's[sd]+'] # s followed by one or more s or d\n \n\nmulti_re_find(test_patterns,test_phrase)", "Searching the phrase using the re check: '[sd]'\n['s', 'd', 's', 'd', 's', 's', 's', 'd', 'd', 'd', 's', 'd', 'd', 'd', 's', 'd', 'd', 'd', 'd', 's', 'd', 's', 'd', 's', 's', 's', 's', 's', 's', 'd', 'd', 'd', 'd']\nSearching the phrase using the re check: 's[sd]+'\n['sdsd', 'sssddd', 'sdddsddd', 'sds', 'sssss', 'sdddd']\n" ] ], [ [ "It makes sense that the first [sd] returns every instance. Also the second input will just return any thing starting with an s in this particular case of the test phrase input.", "_____no_output_____" ], [ "## Exclusion\n\nWe can use ^ to exclude terms by incorporating it into the bracket syntax notation. For example: [^...] will match any single character not in the brackets. Let's see some examples:", "_____no_output_____" ] ], [ [ "test_phrase = 'This is a string! But it has punctuation. How can we remove it?'", "_____no_output_____" ] ], [ [ "Use [^!.? ] to check for matches that are not a !,.,?, or space. Add the + to check that the match appears at least once, this basically translate into finding the words.", "_____no_output_____" ] ], [ [ "re.findall('[^!.? ]+',test_phrase)", "_____no_output_____" ] ], [ [ "## Character Ranges\n\nAs character sets grow larger, typing every character that should (or should not) match could become very tedious. A more compact format using character ranges lets you define a character set to include all of the contiguous characters between a start and stop point. The format used is [start-end].\n\nCommon use cases are to search for a specific range of letters in the alphabet, such [a-f] would return matches with any instance of letters between a and f. \n\nLet's walk through some examples:", "_____no_output_____" ] ], [ [ "\ntest_phrase = 'This is an example sentence. Lets see if we can find some letters.'\n\ntest_patterns=[ '[a-z]+', # sequences of lower case letters\n '[A-Z]+', # sequences of upper case letters\n '[a-zA-Z]+', # sequences of lower or upper case letters\n '[A-Z][a-z]+'] # one upper case letter followed by lower case letters\n \nmulti_re_find(test_patterns,test_phrase)", "Searching the phrase using the re check: '[a-z]+'\n['his', 'is', 'an', 'example', 'sentence', 'ets', 'see', 'if', 'we', 'can', 'find', 'some', 'letters']\nSearching the phrase using the re check: '[A-Z]+'\n['T', 'L']\nSearching the phrase using the re check: '[a-zA-Z]+'\n['This', 'is', 'an', 'example', 'sentence', 'Lets', 'see', 'if', 'we', 'can', 'find', 'some', 'letters']\nSearching the phrase using the re check: '[A-Z][a-z]+'\n['This', 'Lets']\n" ] ], [ [ "## Escape Codes\n\nYou can use special escape codes to find specific types of patterns in your data, such as digits, non-digits,whitespace, and more. For example:\n\n<table border=\"1\" class=\"docutils\">\n<colgroup>\n<col width=\"14%\" />\n<col width=\"86%\" />\n</colgroup>\n<thead valign=\"bottom\">\n<tr class=\"row-odd\"><th class=\"head\">Code</th>\n<th class=\"head\">Meaning</th>\n</tr>\n</thead>\n<tbody valign=\"top\">\n<tr class=\"row-even\"><td><tt class=\"docutils literal\"><span class=\"pre\">\\d</span></tt></td>\n<td>a digit</td>\n</tr>\n<tr class=\"row-odd\"><td><tt class=\"docutils literal\"><span class=\"pre\">\\D</span></tt></td>\n<td>a non-digit</td>\n</tr>\n<tr class=\"row-even\"><td><tt class=\"docutils literal\"><span class=\"pre\">\\s</span></tt></td>\n<td>whitespace (tab, space, newline, etc.)</td>\n</tr>\n<tr class=\"row-odd\"><td><tt class=\"docutils literal\"><span class=\"pre\">\\S</span></tt></td>\n<td>non-whitespace</td>\n</tr>\n<tr class=\"row-even\"><td><tt class=\"docutils literal\"><span class=\"pre\">\\w</span></tt></td>\n<td>alphanumeric</td>\n</tr>\n<tr class=\"row-odd\"><td><tt class=\"docutils literal\"><span class=\"pre\">\\W</span></tt></td>\n<td>non-alphanumeric</td>\n</tr>\n</tbody>\n</table>\n\nEscapes are indicated by prefixing the character with a backslash (\\). Unfortunately, a backslash must itself be escaped in normal Python strings, and that results in expressions that are difficult to read. Using raw strings, created by prefixing the literal value with r, for creating regular expressions eliminates this problem and maintains readability.\n\nPersonally, I think this use of r to escape a backslash is probably one of the things that block someone who is not familiar with regex in Python from being able to read regex code at first. Hopefully after seeing these examples this syntax will become clear.", "_____no_output_____" ] ], [ [ "test_phrase = 'This is a string with some numbers 1233 and a symbol #hashtag'\n\ntest_patterns=[ r'\\d+', # sequence of digits\n r'\\D+', # sequence of non-digits\n r'\\s+', # sequence of whitespace\n r'\\S+', # sequence of non-whitespace\n r'\\w+', # alphanumeric characters\n r'\\W+', # non-alphanumeric\n ]\n\nmulti_re_find(test_patterns,test_phrase)", "Searching the phrase using the re check: '\\\\d+'\n['1233']\nSearching the phrase using the re check: '\\\\D+'\n['This is a string with some numbers ', ' and a symbol #hashtag']\nSearching the phrase using the re check: '\\\\s+'\n[' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ']\nSearching the phrase using the re check: '\\\\S+'\n['This', 'is', 'a', 'string', 'with', 'some', 'numbers', '1233', 'and', 'a', 'symbol', '#hashtag']\nSearching the phrase using the re check: '\\\\w+'\n['This', 'is', 'a', 'string', 'with', 'some', 'numbers', '1233', 'and', 'a', 'symbol', 'hashtag']\nSearching the phrase using the re check: '\\\\W+'\n[' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', ' ', ' #']\n" ] ], [ [ "## Conclusion\n\nYou should now have a solid understanding of how to use the regular expression module in Python. There are a ton of more special character instances, but it would be unreasonable to go through every single use case. Instead take a look at the full [documentation](https://docs.python.org/2.4/lib/re-syntax.html) if you ever need to look up a particular case.\n\nYou can also check out the nice summary tables at this [source](http://www.tutorialspoint.com/python/python_reg_expressions.htm).\n\nGood job!\n", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "You will scrape this <a href=\"https://sandeepmj.github.io/scrape-example-page/homework-site.html\">mockup site</a> that lists a few data points for addiction centers.", "_____no_output_____" ] ], [ [ "pip install icecream", "Collecting icecream\n Downloading icecream-2.1.1-py2.py3-none-any.whl (8.1 kB)\nCollecting colorama>=0.3.9\n Downloading colorama-0.4.4-py2.py3-none-any.whl (16 kB)\nCollecting executing>=0.3.1\n Downloading executing-0.8.1-py2.py3-none-any.whl (16 kB)\nRequirement already satisfied: pygments>=2.2.0 in /usr/local/lib/python3.7/dist-packages (from icecream) (2.6.1)\nCollecting asttokens>=2.0.1\n Downloading asttokens-2.0.5-py2.py3-none-any.whl (20 kB)\nRequirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from asttokens>=2.0.1->icecream) (1.15.0)\nInstalling collected packages: executing, colorama, asttokens, icecream\nSuccessfully installed asttokens-2.0.5 colorama-0.4.4 executing-0.8.1 icecream-2.1.1\n" ], [ "## import library(ies)\nimport requests\nfrom bs4 import BeautifulSoup\nimport pandas as pd\nfrom icecream import ic ", "_____no_output_____" ], [ "## capture the contents of the site in a response object\n\nurl = \"https://sandeepmj.github.io/scrape-example-page/homework-site.html\"\nresponse = requests.get(url)\nic(response.status_code)", "ic| response.status_code: 200\n" ], [ "## generate and print soup\nsoup = BeautifulSoup(response.text, 'html.parser')\nsoup", "_____no_output_____" ], [ "## check data type of soup\ntype(soup)", "_____no_output_____" ], [ "### Return the name of the first center (including the html)\nsoup.find(\"a\")", "_____no_output_____" ], [ "## if there were another a tag that did not have our target, we coud be more specific.\nsoup.find(\"div\", class_=\"wrap\").find(\"h4\").find(\"a\") ", "_____no_output_____" ], [ "### Return only the name of the first center (remove all the html)\nsoup.find(\"a\").get_text()", "_____no_output_____" ], [ "### Return only the URL of the first center\nsoup.find(\"a\").get(\"href\")", "_____no_output_____" ], [ "### Find first instance of ALL a center's data\n### Think of this as the first group of data associated with a company\nsoup.find(\"div\", class_=\"wrap\")", "_____no_output_____" ], [ "#### Find all the instances of every centers' data points.\nsoup.find_all(\"div\", class_=\"wrap\")", "_____no_output_____" ], [ "### Find all the registration data\nsoup.find_all(\"p\", class_=\"registration\")", "_____no_output_____" ] ], [ [ "### Place all the registration data into a list with only the numbers in the format.\nIt should look like this:\n\n```['4235', '4234', '4231']```", "_____no_output_____" ] ], [ [ "## for loop\nregs = soup.find_all(\"p\", class_=\"registration\")\nreg_list_fl = []\nfor item in regs:\n reg_list_fl.append(item.get_text().replace(\"Registration# \", \"\"))\n\nreg_list_fl\n", "_____no_output_____" ], [ "## do it here (create more cells if you need them)\n## via list comprehension\nregs = soup.find_all(\"p\", class_=\"registration\")\nreg_list_lc = [item.get_text().replace(\"Registration# \", \"\") for item in regs]\nreg_list_lc", "_____no_output_____" ] ], [ [ "### Place all the company names into a list.\nIt should look like this:\n\n```['Recovery Foundation','New Horizons','Renewable Light']```", "_____no_output_____" ] ], [ [ "## do it here (create more cells if you need them)\ncos = soup.find_all(\"a\")\ncos", "_____no_output_____" ], [ "### lc\nco_names_list = [item.get_text() for item in cos]\nco_names_list", "_____no_output_____" ] ], [ [ "### Place all the URLS into a list.\n", "_____no_output_____" ] ], [ [ "## do it here (create more cells if you need them)\nco_urls = [item.get(\"href\") for item in cos]\nco_urls", "_____no_output_____" ] ], [ [ "### Place all the status into a list.\nIt should look like this:\n\n```['Passed', 'Failed', 'Passed']```", "_____no_output_____" ] ], [ [ "## do it here (create more cells if you need them)\ncenter_status = soup.find_all(\"p\", class_=\"status\")\ncenter_status", "_____no_output_____" ], [ "status_list = [status.get_text().replace(\"Inspection: \", \"\") for status in center_status ]\nstatus_list", "_____no_output_____" ], [ "", "_____no_output_____" ] ], [ [ "### Turn these lists into dataframes and export to a csv", "_____no_output_____" ] ], [ [ "### use pandas DataFrame method to zip files into a dataframe\ndf = pd.DataFrame(list(zip(co_names_list, reg_list, status_list, co_urls)), \n columns =['center_name', \"registration_number\",'status', 'link'])\n\ndf", "_____no_output_____" ], [ "## export to csv\nfilename = \"recovery_center_list.csv\"\ndf.to_csv(filename, encoding='utf-8', index=False) ## export to csv as utf-8 coding (it just has to be this)\n", "_____no_output_____" ], [ "", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "---\n\n_You are currently looking at **version 1.1** of this notebook. To download notebooks and datafiles, as well as get help on Jupyter notebooks in the Coursera platform, visit the [Jupyter Notebook FAQ](https://www.coursera.org/learn/python-data-analysis/resources/0dhYG) course resource._\n\n---", "_____no_output_____" ], [ "# Assignment 2 - Pandas Introduction\nAll questions are weighted the same in this assignment.\n## Part 1\nThe following code loads the olympics dataset (olympics.csv), which was derrived from the Wikipedia entry on [All Time Olympic Games Medals](https://en.wikipedia.org/wiki/All-time_Olympic_Games_medal_table), and does some basic data cleaning. \n\nThe columns are organized as # of Summer games, Summer medals, # of Winter games, Winter medals, total # number of games, total # of medals. Use this dataset to answer the questions below.", "_____no_output_____" ] ], [ [ "import pandas as pd\n\ndf = pd.read_csv('olympics.csv', index_col=0, skiprows=1)\n\nfor col in df.columns:\n if col[:2]=='01':\n df.rename(columns={col:'Gold'+col[4:]}, inplace=True)\n if col[:2]=='02':\n df.rename(columns={col:'Silver'+col[4:]}, inplace=True)\n if col[:2]=='03':\n df.rename(columns={col:'Bronze'+col[4:]}, inplace=True)\n if col[:1]=='№':\n df.rename(columns={col:'#'+col[1:]}, inplace=True)\n\nnames_ids = df.index.str.split('\\s\\(') # split the index by '('\n\ndf.index = names_ids.str[0] # the [0] element is the country name (new index) \ndf['ID'] = names_ids.str[1].str[:3] # the [1] element is the abbreviation or ID (take first 3 characters from that)\n\ndf = df.drop('Totals')\ndf.head()", "_____no_output_____" ] ], [ [ "### Question 0 (Example)\n\nWhat is the first country in df?\n\n*This function should return a Series.*", "_____no_output_____" ] ], [ [ "# You should write your whole answer within the function provided. The autograder will call\n# this function and compare the return value against the correct solution value\ndef answer_zero():\n # This function returns the row for Afghanistan, which is a Series object. The assignment\n # question description will tell you the general format the autograder is expecting\n return df.iloc[0]\n\n# You can examine what your function returns by calling it in the cell. If you have questions\n# about the assignment formats, check out the discussion forums for any FAQs\nanswer_zero() ", "_____no_output_____" ] ], [ [ "### Question 1\nWhich country has won the most gold medals in summer games?\n\n*This function should return a single string value.*", "_____no_output_____" ] ], [ [ "def answer_one():\n return \"YOUR ANSWER HERE\"", "_____no_output_____" ] ], [ [ "### Question 2\nWhich country had the biggest difference between their summer and winter gold medal counts?\n\n*This function should return a single string value.*", "_____no_output_____" ] ], [ [ "def answer_two():\n return \"YOUR ANSWER HERE\"", "_____no_output_____" ] ], [ [ "### Question 3\nWhich country has the biggest difference between their summer gold medal counts and winter gold medal counts relative to their total gold medal count? \n\n$$\\frac{Summer~Gold - Winter~Gold}{Total~Gold}$$\n\nOnly include countries that have won at least 1 gold in both summer and winter.\n\n*This function should return a single string value.*", "_____no_output_____" ] ], [ [ "def answer_three():\n return \"YOUR ANSWER HERE\"", "_____no_output_____" ] ], [ [ "### Question 4\nWrite a function to update the dataframe to include a new column called \"Points\" which is a weighted value where each gold medal counts for 3 points, silver medals for 2 points, and bronze mdeals for 1 point. The function should return only the column (a Series object) which you created.\n\n*This function should return a Series named `Points` of length 146*", "_____no_output_____" ] ], [ [ "def answer_four():\n return \"YOUR ANSWER HERE\"", "_____no_output_____" ] ], [ [ "## Part 2\nFor the next set of questions, we will be using census data from the [United States Census Bureau](http://www.census.gov/popest/data/counties/totals/2015/CO-EST2015-alldata.html). Counties are political and geographic subdivisions of states in the United States. This dataset contains population data for counties and states in the US from 2010 to 2015. [See this document](http://www.census.gov/popest/data/counties/totals/2015/files/CO-EST2015-alldata.pdf) for a description of the variable names.\n\nThe census dataset (census.csv) should be loaded as census_df. Answer questions using this as appropriate.\n\n### Question 5\nWhich state has the most counties in it? (hint: consider the sumlevel key carefully! You'll need this for future questions too...)\n\n*This function should return a single string value.*", "_____no_output_____" ] ], [ [ "census_df = pd.read_csv('census.csv')\ncensus_df.head()", "_____no_output_____" ], [ "def answer_five():\n return \"YOUR ANSWER HERE\"", "_____no_output_____" ] ], [ [ "### Question 6\nOnly looking at the three most populous counties for each state, what are the three most populous states (in order of highest population to lowest population)?\n\n*This function should return a list of string values.*", "_____no_output_____" ] ], [ [ "def answer_six():\n return \"YOUR ANSWER HERE\"", "_____no_output_____" ] ], [ [ "### Question 7\nWhich county has had the largest absolute change in population within the period 2010-2015? (Hint: population values are stored in columns POPESTIMATE2010 through POPESTIMATE2015, you need to consider all six columns.)\n\ne.g. If County Population in the 5 year period is 100, 120, 80, 105, 100, 130, then its largest change in the period would be |130-80| = 50.\n\n*This function should return a single string value.*", "_____no_output_____" ] ], [ [ "def answer_seven():\n return \"YOUR ANSWER HERE\"", "_____no_output_____" ] ], [ [ "### Question 8\nIn this datafile, the United States is broken up into four regions using the \"REGION\" column. \n\nCreate a query that finds the counties that belong to regions 1 or 2, whose name starts with 'Washington', and whose POPESTIMATE2015 was greater than their POPESTIMATE 2014.\n\n*This function should return a 5x2 DataFrame with the columns = ['STNAME', 'CTYNAME'] and the same index ID as the census_df (sorted ascending by index).*", "_____no_output_____" ] ], [ [ "def answer_eight():\n return \"YOUR ANSWER HERE\"", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "car=\"car1.jpeg\"", "_____no_output_____" ], [ "import cv2\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport matplotlib.image as mpimg", "_____no_output_____" ], [ "car1=mpimg.imread('car1.jpeg')", "_____no_output_____" ], [ "car1.shape", "_____no_output_____" ], [ "type(car1)", "_____no_output_____" ], [ "plt.imshow(car1)", "_____no_output_____" ], [ "#plt.imshow(car)", "_____no_output_____" ], [ "car1_cv2=cv2.imread('car1.jpeg')\ncolor_car_cv2=cv2.cvtColor(car1_cv2, cv2.COLOR_BGR2RGB)", "_____no_output_____" ], [ "plt.imshow(color_car_cv2)", "_____no_output_____" ], [ "car1_cv2_BGR_Gray=cv2.cvtColor(car1,cv2.COLOR_BGR2GRAY)\nplt.imshow(car1_cv2_BGR_Gray, cmap=\"gray\")", "_____no_output_____" ], [ "car1_cv2_BGR_Gray.shape", "_____no_output_____" ], [ "car1_cv2_BGR_Gray.min(), car1_cv2_BGR_Gray.max()", "_____no_output_____" ], [ "car1_cv2_BGR_Gray[0][0]", "_____no_output_____" ], [ "cv2.imwrite(\"car_gray.jpeg\",car1_cv2_BGR_Gray)", "_____no_output_____" ], [ "car1_cv2_BGR_Gray[0][200]", "_____no_output_____" ] ], [ [ "# Visualize all the RGB channel", "_____no_output_____" ] ], [ [ "def visualize_RGB_Channels(imgArray=None, fig_size=(10,7)):\n # spliting the RGB components\n B,G,R=cv2.split(imgArray)\n #zero matrix\n Z=np.zeros(B.shape,dtype=B.dtype)\n #initilize subplot\n fig,ax=plt.subplots(2,2, figsize=fig_size)\n [axi.set_axis_off() for axi in ax.ravel()]\n ax[0,0].set_title(\"Original image\")\n ax[0,0].imshow(cv2.merge((R,G,B)))\n ax[0,1].set_title(\"Red Chanel\")\n ax[0,1].imshow(cv2.merge((R,Z,Z)))\n ax[1,0].set_title(\"Green Chanel\")\n ax[1,0].imshow(cv2.merge((Z,G,Z)))\n ax[1,1].set_title(\"Blue Chanel\")\n ax[1,1].imshow(cv2.merge((Z,Z,B)))", "_____no_output_____" ], [ "visualize_RGB_Channels(imgArray=car1_cv2)", "_____no_output_____" ], [ "random_colored_image=np.random.randint(0,255,(6,6,3))\nrandom_colored_image.shape", "_____no_output_____" ], [ "visualize_RGB_Channels(imgArray=random_colored_image)", "_____no_output_____" ], [ "random_colored_image[0,0:]", "_____no_output_____" ], [ "random_colored_image[0,0,:]", "_____no_output_____" ], [ "random_colored_image[-1,-1,:]", "_____no_output_____" ] ], [ [ "# Filters", "_____no_output_____" ] ], [ [ "sobel=np.array([[1,0,-1],[2,0,-2],[1,0,-1]])\nprint(sobel)", "[[ 1 0 -1]\n [ 2 0 -2]\n [ 1 0 -1]]\n" ], [ "sobel.T", "_____no_output_____" ], [ "example1=[[0,0,0,255,255,255],\n [0,0,0,255,255,255],\n [0,0,0,255,255,255],\n [0,0,0,255,255,255],\n [0,0,0,255,255,255],\n [0,0,0,255,255,255]]", "_____no_output_____" ], [ "example1=np.array(example1)", "_____no_output_____" ], [ "plt.imshow(example1, cmap=\"gray\")", "_____no_output_____" ] ], [ [ "# Apply filter on this image", "_____no_output_____" ] ], [ [ "def find_edges(imgFilter=None, picture=None):\n # extract row and column of an input picture\n p_row,p_col=picture.shape\n k=imgFilter.shape[0]\n temp=list()\n strides=1\n #resultant rows and columns\n final_columns=(p_col -k)//strides +1\n \n final_rows=(p_row -k)//strides +1\n #take vertically down sr´tride accross row by row\n for v_stride in range(final_rows):\n for h_stride in range(final_columns):\n target_area_of_pic=picture[v_stride:v_stride +k, h_stride:h_stride+k]\n temp.append(sum(sum(imgFilter* target_area_of_pic)))\n #print(temp)\n return np.array(temp).reshape(final_rows,final_columns)\n ", "_____no_output_____" ], [ "result=find_edges(sobel, example1)\nresult\nsum(sum(result))", "[0, -1020, -1020, 0, 0, -1020, -1020, 0, 0, -1020, -1020, 0, 0, -1020, -1020, 0]\n" ], [ "example1", "_____no_output_____" ], [ "sum(example1)", "_____no_output_____" ], [ "sum(sum(example1))", "_____no_output_____" ], [ "plt.imshow(result, cmap=\"gray\")", "_____no_output_____" ], [ "sobel.T", "_____no_output_____" ], [ "result_T=find_edges(sobel.T, example1)\nresult_T", "[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]\n" ], [ "plt.imshow(result_T, cmap=\"gray\")", "_____no_output_____" ], [ "example1", "_____no_output_____" ], [ "example_T=example1.T\nplt.imshow(example_T,cmap=\"gray\")", "_____no_output_____" ], [ "result_T1=find_edges(sobel.T, example_T)\nresult_T1", "[0, 0, 0, 0, -1020, -1020, -1020, -1020, -1020, -1020, -1020, -1020, 0, 0, 0, 0]\n" ], [ "plt.imshow(result_T1, cmap=\"gray\")", "_____no_output_____" ], [ "result_car=find_edges(sobel,car1_cv2_BGR_Gray)\nresult_car", "_____no_output_____" ], [ "plt.imshow(result_car, cmap=\"gray\")", "_____no_output_____" ] ], [ [ "# lets now apply horizontal edges", "_____no_output_____" ] ], [ [ "result_car_hor=find_edges(sobel.T, car1_cv2_BGR_Gray)", "_____no_output_____" ], [ "plt.imshow(result_car, cmap=\"gray\")", "_____no_output_____" ], [ "example1", "_____no_output_____" ], [ "example1=[[255,0,0,0,255,255,255,255,0,0,0,255],\n [0,0,0,0,255,255,255,255,0,0,0,0],\n [0,0,0,0,255,255,255,255,255,255,255,255],\n [0,0,0,0,255,255,255,255,255,255,255,255],\n [0,0,0,0,255,255,255,255,255,255,255,255],\n [0,0,0,0,255,255,255,255,255,255,255,255],\n [0,0,0,0,255,255,255,255,255,255,255,255],\n [0,0,0,0,255,255,255,255,0,0,0,0],\n [0,0,0,0,255,255,255,255,0,0,0,0],\n [255,0,0,0,255,255,255,255,0,0,0,255]]", "_____no_output_____" ], [ "example1=np.array(example1)", "_____no_output_____" ], [ "plt.imshow(example1, cmap=\"gray\")", "_____no_output_____" ], [ "result=find_edges(sobel.T, example1)\nplt.imshow(result, cmap=\"gray\")", "_____no_output_____" ], [ "result.shape", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "pip install nltk", "Requirement already satisfied: nltk in c:\\python38\\lib\\site-packages (3.5)Note: you may need to restart the kernel to use updated packages.\nWARNING: You are using pip version 20.2.2; however, version 21.0.1 is available.\nYou should consider upgrading via the 'C:\\Python38\\python.exe -m pip install --upgrade pip' command.\n\nRequirement already satisfied: click in c:\\python38\\lib\\site-packages (from nltk) (7.1.2)\nRequirement already satisfied: joblib in c:\\python38\\lib\\site-packages (from nltk) (1.0.1)\nRequirement already satisfied: regex in c:\\python38\\lib\\site-packages (from nltk) (2021.3.17)\nRequirement already satisfied: tqdm in c:\\python38\\lib\\site-packages (from nltk) (4.59.0)\n" ], [ "import nltk\nimport string\nimport re", "_____no_output_____" ], [ "texto_original = \"\"\"Algoritmos inteligentes de aprendizados correndo supervisionados utilizam dados coletados. A partir dos dados coletados, um conjunto de característica é extraído. As características podem ser estruturais ou estatísticas. Correr correste corrida inteligente. As característica estruturais estabelecem relações entre os dados, inteligência enquanto que as estatísticas são características quantitativas. A partir das corrido características, os modelos de aprendizado de máquina corremos são construídos para o reconhecimento de atividades humanas.\"\"\"", "_____no_output_____" ], [ "texto_original", "_____no_output_____" ], [ "#texto_original = re.sub(r'\\s+', '', texto_original)\n#texto_original", "_____no_output_____" ], [ "def converte_para_minusculas(texto):\n texto_formatado = texto.lower()\n return texto_formatado", "_____no_output_____" ], [ "texto_formatado = converte_para_minusculas(texto_original)\ntexto_formatado", "_____no_output_____" ], [ "nltk.download('stopwords')", "[nltk_data] Downloading package stopwords to\n[nltk_data] C:\\Users\\andre\\AppData\\Roaming\\nltk_data...\n[nltk_data] Package stopwords is already up-to-date!\n" ], [ "stopwords = nltk.corpus.stopwords.words('portuguese')", "_____no_output_____" ], [ "print(stopwords)", "['de', 'a', 'o', 'que', 'e', 'é', 'do', 'da', 'em', 'um', 'para', 'com', 'não', 'uma', 'os', 'no', 'se', 'na', 'por', 'mais', 'as', 'dos', 'como', 'mas', 'ao', 'ele', 'das', 'à', 'seu', 'sua', 'ou', 'quando', 'muito', 'nos', 'já', 'eu', 'também', 'só', 'pelo', 'pela', 'até', 'isso', 'ela', 'entre', 'depois', 'sem', 'mesmo', 'aos', 'seus', 'quem', 'nas', 'me', 'esse', 'eles', 'você', 'essa', 'num', 'nem', 'suas', 'meu', 'às', 'minha', 'numa', 'pelos', 'elas', 'qual', 'nós', 'lhe', 'deles', 'essas', 'esses', 'pelas', 'este', 'dele', 'tu', 'te', 'vocês', 'vos', 'lhes', 'meus', 'minhas', 'teu', 'tua', 'teus', 'tuas', 'nosso', 'nossa', 'nossos', 'nossas', 'dela', 'delas', 'esta', 'estes', 'estas', 'aquele', 'aquela', 'aqueles', 'aquelas', 'isto', 'aquilo', 'estou', 'está', 'estamos', 'estão', 'estive', 'esteve', 'estivemos', 'estiveram', 'estava', 'estávamos', 'estavam', 'estivera', 'estivéramos', 'esteja', 'estejamos', 'estejam', 'estivesse', 'estivéssemos', 'estivessem', 'estiver', 'estivermos', 'estiverem', 'hei', 'há', 'havemos', 'hão', 'houve', 'houvemos', 'houveram', 'houvera', 'houvéramos', 'haja', 'hajamos', 'hajam', 'houvesse', 'houvéssemos', 'houvessem', 'houver', 'houvermos', 'houverem', 'houverei', 'houverá', 'houveremos', 'houverão', 'houveria', 'houveríamos', 'houveriam', 'sou', 'somos', 'são', 'era', 'éramos', 'eram', 'fui', 'foi', 'fomos', 'foram', 'fora', 'fôramos', 'seja', 'sejamos', 'sejam', 'fosse', 'fôssemos', 'fossem', 'for', 'formos', 'forem', 'serei', 'será', 'seremos', 'serão', 'seria', 'seríamos', 'seriam', 'tenho', 'tem', 'temos', 'tém', 'tinha', 'tínhamos', 'tinham', 'tive', 'teve', 'tivemos', 'tiveram', 'tivera', 'tivéramos', 'tenha', 'tenhamos', 'tenham', 'tivesse', 'tivéssemos', 'tivessem', 'tiver', 'tivermos', 'tiverem', 'terei', 'terá', 'teremos', 'terão', 'teria', 'teríamos', 'teriam']\n" ], [ "len(stopwords)", "_____no_output_____" ], [ "stopwords.append('ola')\nprint(stopwords)\nstopwords.remove('ola')\nprint(stopwords)", "['de', 'a', 'o', 'que', 'e', 'é', 'do', 'da', 'em', 'um', 'para', 'com', 'não', 'uma', 'os', 'no', 'se', 'na', 'por', 'mais', 'as', 'dos', 'como', 'mas', 'ao', 'ele', 'das', 'à', 'seu', 'sua', 'ou', 'quando', 'muito', 'nos', 'já', 'eu', 'também', 'só', 'pelo', 'pela', 'até', 'isso', 'ela', 'entre', 'depois', 'sem', 'mesmo', 'aos', 'seus', 'quem', 'nas', 'me', 'esse', 'eles', 'você', 'essa', 'num', 'nem', 'suas', 'meu', 'às', 'minha', 'numa', 'pelos', 'elas', 'qual', 'nós', 'lhe', 'deles', 'essas', 'esses', 'pelas', 'este', 'dele', 'tu', 'te', 'vocês', 'vos', 'lhes', 'meus', 'minhas', 'teu', 'tua', 'teus', 'tuas', 'nosso', 'nossa', 'nossos', 'nossas', 'dela', 'delas', 'esta', 'estes', 'estas', 'aquele', 'aquela', 'aqueles', 'aquelas', 'isto', 'aquilo', 'estou', 'está', 'estamos', 'estão', 'estive', 'esteve', 'estivemos', 'estiveram', 'estava', 'estávamos', 'estavam', 'estivera', 'estivéramos', 'esteja', 'estejamos', 'estejam', 'estivesse', 'estivéssemos', 'estivessem', 'estiver', 'estivermos', 'estiverem', 'hei', 'há', 'havemos', 'hão', 'houve', 'houvemos', 'houveram', 'houvera', 'houvéramos', 'haja', 'hajamos', 'hajam', 'houvesse', 'houvéssemos', 'houvessem', 'houver', 'houvermos', 'houverem', 'houverei', 'houverá', 'houveremos', 'houverão', 'houveria', 'houveríamos', 'houveriam', 'sou', 'somos', 'são', 'era', 'éramos', 'eram', 'fui', 'foi', 'fomos', 'foram', 'fora', 'fôramos', 'seja', 'sejamos', 'sejam', 'fosse', 'fôssemos', 'fossem', 'for', 'formos', 'forem', 'serei', 'será', 'seremos', 'serão', 'seria', 'seríamos', 'seriam', 'tenho', 'tem', 'temos', 'tém', 'tinha', 'tínhamos', 'tinham', 'tive', 'teve', 'tivemos', 'tiveram', 'tivera', 'tivéramos', 'tenha', 'tenhamos', 'tenham', 'tivesse', 'tivéssemos', 'tivessem', 'tiver', 'tivermos', 'tiverem', 'terei', 'terá', 'teremos', 'terão', 'teria', 'teríamos', 'teriam', 'ola']\n['de', 'a', 'o', 'que', 'e', 'é', 'do', 'da', 'em', 'um', 'para', 'com', 'não', 'uma', 'os', 'no', 'se', 'na', 'por', 'mais', 'as', 'dos', 'como', 'mas', 'ao', 'ele', 'das', 'à', 'seu', 'sua', 'ou', 'quando', 'muito', 'nos', 'já', 'eu', 'também', 'só', 'pelo', 'pela', 'até', 'isso', 'ela', 'entre', 'depois', 'sem', 'mesmo', 'aos', 'seus', 'quem', 'nas', 'me', 'esse', 'eles', 'você', 'essa', 'num', 'nem', 'suas', 'meu', 'às', 'minha', 'numa', 'pelos', 'elas', 'qual', 'nós', 'lhe', 'deles', 'essas', 'esses', 'pelas', 'este', 'dele', 'tu', 'te', 'vocês', 'vos', 'lhes', 'meus', 'minhas', 'teu', 'tua', 'teus', 'tuas', 'nosso', 'nossa', 'nossos', 'nossas', 'dela', 'delas', 'esta', 'estes', 'estas', 'aquele', 'aquela', 'aqueles', 'aquelas', 'isto', 'aquilo', 'estou', 'está', 'estamos', 'estão', 'estive', 'esteve', 'estivemos', 'estiveram', 'estava', 'estávamos', 'estavam', 'estivera', 'estivéramos', 'esteja', 'estejamos', 'estejam', 'estivesse', 'estivéssemos', 'estivessem', 'estiver', 'estivermos', 'estiverem', 'hei', 'há', 'havemos', 'hão', 'houve', 'houvemos', 'houveram', 'houvera', 'houvéramos', 'haja', 'hajamos', 'hajam', 'houvesse', 'houvéssemos', 'houvessem', 'houver', 'houvermos', 'houverem', 'houverei', 'houverá', 'houveremos', 'houverão', 'houveria', 'houveríamos', 'houveriam', 'sou', 'somos', 'são', 'era', 'éramos', 'eram', 'fui', 'foi', 'fomos', 'foram', 'fora', 'fôramos', 'seja', 'sejamos', 'sejam', 'fosse', 'fôssemos', 'fossem', 'for', 'formos', 'forem', 'serei', 'será', 'seremos', 'serão', 'seria', 'seríamos', 'seriam', 'tenho', 'tem', 'temos', 'tém', 'tinha', 'tínhamos', 'tinham', 'tive', 'teve', 'tivemos', 'tiveram', 'tivera', 'tivéramos', 'tenha', 'tenhamos', 'tenham', 'tivesse', 'tivéssemos', 'tivessem', 'tiver', 'tivermos', 'tiverem', 'terei', 'terá', 'teremos', 'terão', 'teria', 'teríamos', 'teriam']\n" ], [ "def remove_stopwords(texto):\n texto_formatado = converte_para_minusculas(texto)\n tokens = [] # Lista de tokens\n for token in nltk.word_tokenize(texto_formatado): # token é cada uma das palavras\n tokens.append(token) # Insere na lista de tokens\n \n # Pega apenas os tokens que não estão nas stopwords\n tokens = [cada_token for cada_token in tokens if cada_token not in stopwords and cada_token not in string.punctuation]\n\n texto_formatado = ' '.join([str(cada_token) for cada_token in tokens if not cada_token.isdigit()])\n\n return texto_formatado", "_____no_output_____" ], [ "texto_formatado = remove_stopwords(texto_original)", "_____no_output_____" ], [ "string.punctuation", "_____no_output_____" ], [ "frequencia_palavras = nltk.FreqDist(nltk.word_tokenize(texto_formatado))\nfrequencia_palavras", "_____no_output_____" ], [ "frequencia_palavras.keys()", "_____no_output_____" ], [ "frequencia_maxima = max(frequencia_palavras.values())\nfrequencia_maxima", "_____no_output_____" ], [ "for cada_palavra in frequencia_palavras.keys():\n frequencia_palavras[cada_palavra] = frequencia_palavras[cada_palavra]/frequencia_maxima", "_____no_output_____" ], [ "frequencia_palavras", "_____no_output_____" ], [ "sentencas = nltk.sent_tokenize(texto_original)\nsentencas", "_____no_output_____" ], [ "notas_das_sentencas = {}\nfor cada_sentenca in sentencas:\n # print(cada_sentenca)\n for cada_palavra in nltk.word_tokenize(cada_sentenca.lower()):\n # print(cada_palavra)\n if cada_palavra in frequencia_palavras.keys():\n if cada_sentenca not in notas_das_sentencas:\n notas_das_sentencas[cada_sentenca] = frequencia_palavras[cada_palavra]\n else:\n notas_das_sentencas[cada_sentenca] += frequencia_palavras[cada_palavra]", "_____no_output_____" ], [ "notas_das_sentencas", "_____no_output_____" ], [ "import heapq", "_____no_output_____" ], [ "melhores_sentencas = heapq.nlargest(3, notas_das_sentencas, key = notas_das_sentencas.get)\nmelhores_sentencas", "_____no_output_____" ], [ "resumo = ''.join(melhores_sentencas)\nresumo", "_____no_output_____" ], [ "from IPython.core.display import HTML", "_____no_output_____" ], [ "texto_final = ''\ndisplay(HTML(f'<h1> RESUMO GERADO AUTOMATICAMENTE </h1>'))\n\nfor cada_sentenca in sentencas:\n #texto_final = ''\n #texto_final += cada_sentenca\n if cada_sentenca in melhores_sentencas:\n texto_final += str(cada_sentenca).replace(cada_sentenca, f\"<mark>{cada_sentenca}</mark>\")\n else:\n texto_final += str(cada_sentenca)\n texto_final += ' '\n\ndisplay(HTML(f\"\"\"{texto_final}\"\"\"))", "_____no_output_____" ], [ "pip install goose3", "Requirement already satisfied: goose3 in c:\\python38\\lib\\site-packages (3.1.8)\nRequirement already satisfied: Pillow in c:\\python38\\lib\\site-packages (from goose3) (8.1.0)\nRequirement already satisfied: jieba in c:\\python38\\lib\\site-packages (from goose3) (0.42.1)\nRequirement already satisfied: beautifulsoup4 in c:\\python38\\lib\\site-packages (from goose3) (4.9.1)\nNote: you may need to restart the kernel to use updated packages.\nWARNING: You are using pip version 20.2.2; however, version 21.0.1 is available.\nYou should consider upgrading via the 'C:\\Python38\\python.exe -m pip install --upgrade pip' command.\nRequirement already satisfied: nltk in c:\\python38\\lib\\site-packages (from goose3) (3.5)\nRequirement already satisfied: requests in c:\\python38\\lib\\site-packages (from goose3) (2.24.0)\nRequirement already satisfied: lxml in c:\\python38\\lib\\site-packages (from goose3) (4.6.3)\nRequirement already satisfied: python-dateutil in c:\\users\\andre\\appdata\\roaming\\python\\python38\\site-packages (from goose3) (2.8.1)\nRequirement already satisfied: cssselect in c:\\python38\\lib\\site-packages (from goose3) (1.1.0)\nRequirement already satisfied: soupsieve>1.2 in c:\\python38\\lib\\site-packages (from beautifulsoup4->goose3) (2.0.1)\nRequirement already satisfied: click in c:\\python38\\lib\\site-packages (from nltk->goose3) (7.1.2)\nRequirement already satisfied: joblib in c:\\python38\\lib\\site-packages (from nltk->goose3) (1.0.1)\nRequirement already satisfied: regex in c:\\python38\\lib\\site-packages (from nltk->goose3) (2021.3.17)\nRequirement already satisfied: tqdm in c:\\python38\\lib\\site-packages (from nltk->goose3) (4.59.0)\nRequirement already satisfied: chardet<4,>=3.0.2 in c:\\python38\\lib\\site-packages (from requests->goose3) (3.0.4)\nRequirement already satisfied: certifi>=2017.4.17 in c:\\python38\\lib\\site-packages (from requests->goose3) (2020.6.20)\nRequirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in c:\\python38\\lib\\site-packages (from requests->goose3) (1.25.10)\nRequirement already satisfied: idna<3,>=2.5 in c:\\python38\\lib\\site-packages (from requests->goose3) (2.10)\nRequirement already satisfied: six>=1.5 in c:\\users\\andre\\appdata\\roaming\\python\\python38\\site-packages (from python-dateutil->goose3) (1.15.0)\n" ] ], [ [ "# Tarefa 1", "_____no_output_____" ], [ "## 1. Stemizacao", "_____no_output_____" ] ], [ [ "from nltk.stem.snowball import SnowballStemmer\n\n# É importante definir a lingua\nstemizador = SnowballStemmer('portuguese')", "_____no_output_____" ], [ "palavras_stemizadas = []\nfor palavra in nltk.word_tokenize(texto_formatado):\n print(palavra, ' = ', stemizador.stem(palavra))\n palavras_stemizadas.append(stemizador.stem(palavra))", "algoritmos = algoritm\ninteligentes = inteligent\naprendizados = aprendiz\ncorrendo = corr\nsupervisionados = supervision\nutilizam = utiliz\ndados = dad\ncoletados = colet\npartir = part\ndados = dad\ncoletados = colet\nconjunto = conjunt\ncaracterística = característ\nextraído = extraíd\ncaracterísticas = característ\npodem = pod\nser = ser\nestruturais = estrutur\nestatísticas = estatíst\ncorrer = corr\ncorreste = corr\ncorrida = corr\ninteligente = inteligent\ncaracterística = característ\nestruturais = estrutur\nestabelecem = estabelec\nrelações = relaçõ\ndados = dad\ninteligência = inteligent\nenquanto = enquant\nestatísticas = estatíst\ncaracterísticas = característ\nquantitativas = quantit\npartir = part\ncorrido = corr\ncaracterísticas = característ\nmodelos = model\naprendizado = aprendiz\nmáquina = máquin\ncorremos = corr\nconstruídos = construíd\nreconhecimento = reconhec\natividades = ativ\nhumanas = human\n" ], [ "print(palavras_stemizadas)", "['algoritm', 'inteligent', 'aprendiz', 'corr', 'supervision', 'utiliz', 'dad', 'colet', 'part', 'dad', 'colet', 'conjunt', 'característ', 'extraíd', 'característ', 'pod', 'ser', 'estrutur', 'estatíst', 'corr', 'corr', 'corr', 'inteligent', 'característ', 'estrutur', 'estabelec', 'relaçõ', 'dad', 'inteligent', 'enquant', 'estatíst', 'característ', 'quantit', 'part', 'corr', 'característ', 'model', 'aprendiz', 'máquin', 'corr', 'construíd', 'reconhec', 'ativ', 'human']\n" ], [ "resultado = ' '.join([str(cada_token) for cada_token in palavras_stemizadas if not cada_token.isdigit()])\nprint(resultado)", "algoritm inteligent aprendiz corr supervision utiliz dad colet part dad colet conjunt característ extraíd característ pod ser estrutur estatíst corr corr corr inteligent característ estrutur estabelec relaçõ dad inteligent enquant estatíst característ quantit part corr característ model aprendiz máquin corr construíd reconhec ativ human\n" ], [ "frequencia_palavras = nltk.FreqDist(nltk.word_tokenize(resultado))\nfrequencia_palavras", "_____no_output_____" ], [ "frequencia_maxima = max(frequencia_palavras.values())\nfrequencia_maxima", "_____no_output_____" ], [ "for cada_palavra in frequencia_palavras.keys():\n frequencia_palavras[cada_palavra] = frequencia_palavras[cada_palavra]/frequencia_maxima", "_____no_output_____" ], [ "frequencia_palavras", "_____no_output_____" ], [ "notas_das_sentencas = {}\nfor cada_sentenca in sentencas:\n # print(cada_sentenca)\n for cada_palavra in nltk.word_tokenize(cada_sentenca.lower()):\n # print(cada_palavra)\n aux = stemizador.stem(cada_palavra)\n if aux in frequencia_palavras.keys():\n if cada_sentenca not in notas_das_sentencas:\n notas_das_sentencas[cada_sentenca] = frequencia_palavras[aux]\n else:\n notas_das_sentencas[cada_sentenca] += frequencia_palavras[aux]", "_____no_output_____" ], [ "notas_das_sentencas", "_____no_output_____" ], [ "melhores_sentencas = heapq.nlargest(3, notas_das_sentencas, key = notas_das_sentencas.get)\nmelhores_sentencas", "_____no_output_____" ], [ "resumo = ''.join(melhores_sentencas)\nresumo", "_____no_output_____" ], [ "texto_final = ''\ndisplay(HTML(f'<h1> RESUMO GERADO AUTOMATICAMENTE (Stemizadas)</h1>'))\n\nfor cada_sentenca in sentencas:\n #texto_final = ''\n #texto_final += cada_sentenca\n if cada_sentenca in melhores_sentencas:\n texto_final += str(cada_sentenca).replace(cada_sentenca, f\"<mark>{cada_sentenca}</mark>\")\n else:\n texto_final += str(cada_sentenca)\n texto_final += ' '\n\ndisplay(HTML(f\"\"\"{texto_final}\"\"\"))", "_____no_output_____" ] ], [ [ "## 2. Lematizacao", "_____no_output_____" ] ], [ [ "import spacy", "_____no_output_____" ], [ "!python -m spacy download pt_core_news_sm", "Requirement already satisfied: pt-core-news-sm==3.0.0 from https://github.com/explosion/spacy-models/releases/download/pt_core_news_sm-3.0.0/pt_core_news_sm-3.0.0-py3-none-any.whl#egg=pt_core_news_sm==3.0.0 in c:\\python38\\lib\\site-packages (3.0.0)\nRequirement already satisfied: spacy<3.1.0,>=3.0.0 in c:\\python38\\lib\\site-packages (from pt-core-news-sm==3.0.0) (3.0.5)\nRequirement already satisfied: wasabi<1.1.0,>=0.8.1 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (0.8.2)\nRequirement already satisfied: requests<3.0.0,>=2.13.0 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (2.24.0)\nRequirement already satisfied: pydantic<1.8.0,>=1.7.1 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (1.7.3)\nRequirement already satisfied: preshed<3.1.0,>=3.0.2 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (3.0.5)\nRequirement already satisfied: typer<0.4.0,>=0.3.0 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (0.3.2)\nRequirement already satisfied: jinja2 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (2.11.3)\nRequirement already satisfied: setuptools in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (47.1.0)\nRequirement already satisfied: thinc<8.1.0,>=8.0.2 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (8.0.2)\nRequirement already satisfied: packaging>=20.0 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (20.9)\nRequirement already satisfied: catalogue<2.1.0,>=2.0.1 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (2.0.1)\nRequirement already satisfied: srsly<3.0.0,>=2.4.0 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (2.4.0)\nRequirement already satisfied: cymem<2.1.0,>=2.0.2 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (2.0.5)\nRequirement already satisfied: pathy>=0.3.5 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (0.4.0)\nRequirement already satisfied: spacy-legacy<3.1.0,>=3.0.0 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (3.0.1)\nRequirement already satisfied: blis<0.8.0,>=0.4.0 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (0.7.4)\nRequirement already satisfied: tqdm<5.0.0,>=4.38.0 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (4.59.0)\nRequirement already satisfied: murmurhash<1.1.0,>=0.28.0 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (1.0.5)\nRequirement already satisfied: numpy>=1.15.0 in c:\\python38\\lib\\site-packages (from spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (1.20.1)\nRequirement already satisfied: idna<3,>=2.5 in c:\\python38\\lib\\site-packages (from requests<3.0.0,>=2.13.0->spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (2.10)\nRequirement already satisfied: chardet<4,>=3.0.2 in c:\\python38\\lib\\site-packages (from requests<3.0.0,>=2.13.0->spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (3.0.4)\nRequirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in c:\\python38\\lib\\site-packages (from requests<3.0.0,>=2.13.0->spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (1.25.10)\nRequirement already satisfied: certifi>=2017.4.17 in c:\\python38\\lib\\site-packages (from requests<3.0.0,>=2.13.0->spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (2020.6.20)\nRequirement already satisfied: click<7.2.0,>=7.1.1 in c:\\python38\\lib\\site-packages (from typer<0.4.0,>=0.3.0->spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (7.1.2)\nRequirement already satisfied: MarkupSafe>=0.23 in c:\\python38\\lib\\site-packages (from jinja2->spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (1.1.1)\nRequirement already satisfied: pyparsing>=2.0.2 in c:\\python38\\lib\\site-packages (from packaging>=20.0->spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (2.4.7)\nRequirement already satisfied: smart-open<4.0.0,>=2.2.0 in c:\\python38\\lib\\site-packages (from pathy>=0.3.5->spacy<3.1.0,>=3.0.0->pt-core-news-sm==3.0.0) (3.0.0)\n✔ Download and installation successful\nYou can now load the package via spacy.load('pt_core_news_sm')\nWARNING: You are using pip version 20.2.2; however, version 21.0.1 is available.\nYou should consider upgrading via the 'C:\\Python38\\python.exe -m pip install --upgrade pip' command.\n" ], [ "pln = spacy.load('pt_core_news_sm')\npln", "_____no_output_____" ], [ "palavras = pln(texto_formatado)", "_____no_output_____" ], [ "# Spacy já separa as palavras em tokens\npalavras_lematizadas = []\nfor palavra in palavras:\n #print(palavra.text, ' = ', palavra.lemma_)\n palavras_lematizadas.append(palavra.lemma_)\n\nprint(palavras_lematizadas)", "['algoritmo', 'inteligente', 'aprendizado', 'correr', 'supervisionar', 'utilizar', 'dar', 'coletados', 'partir', 'dar', 'coletados', 'conjuntar', 'característico', 'extrair', 'característico', 'poder', 'ser', 'estruturar', 'estatístico', 'correr', 'correr', 'corrido', 'inteligente', 'característico', 'estruturar', 'estabelecer', 'relação', 'dar', 'inteligência', 'enquanto', 'estatístico', 'característico', 'quantitativo', 'partir', 'correr', 'característico', 'modelo', 'aprendizado', 'máquina', 'correr', 'construir', 'reconhecimento', 'atividades', 'humano']\n" ], [ "resultado = ' '.join([str(cada_token) for cada_token in palavras_lematizadas if not cada_token.isdigit()])\nprint(resultado)", "algoritmo inteligente aprendizado correr supervisionar utilizar dar coletados partir dar coletados conjuntar característico extrair característico poder ser estruturar estatístico correr correr corrido inteligente característico estruturar estabelecer relação dar inteligência enquanto estatístico característico quantitativo partir correr característico modelo aprendizado máquina correr construir reconhecimento atividades humano\n" ], [ "frequencia_palavras = nltk.FreqDist(nltk.word_tokenize(resultado))\nfrequencia_palavras", "_____no_output_____" ], [ "frequencia_maxima = max(frequencia_palavras.values())\nfrequencia_maxima", "_____no_output_____" ], [ "for cada_palavra in frequencia_palavras.keys():\n frequencia_palavras[cada_palavra] = frequencia_palavras[cada_palavra]/frequencia_maxima\n\nfrequencia_palavras", "_____no_output_____" ], [ "notas_das_sentencas = {}\nfor cada_sentenca in sentencas:\n # print(cada_sentenca)\n \n for cada_palavra in pln(cada_sentenca):\n # print(cada_palavra)\n aux = cada_palavra.lemma_\n if aux in frequencia_palavras.keys():\n if cada_sentenca not in notas_das_sentencas:\n notas_das_sentencas[cada_sentenca] = frequencia_palavras[aux]\n else:\n notas_das_sentencas[cada_sentenca] += frequencia_palavras[aux]", "_____no_output_____" ], [ "notas_das_sentencas", "_____no_output_____" ], [ "melhores_sentencas = heapq.nlargest(3, notas_das_sentencas, key = notas_das_sentencas.get)\nmelhores_sentencas", "_____no_output_____" ], [ "resumo = ''.join(melhores_sentencas)\nresumo", "_____no_output_____" ], [ "texto_final = ''\ndisplay(HTML(f'<h1> RESUMO GERADO AUTOMATICAMENTE (Lematizacao)</h1>'))\n\nfor cada_sentenca in sentencas:\n #texto_final = ''\n #texto_final += cada_sentenca\n if cada_sentenca in melhores_sentencas:\n texto_final += str(cada_sentenca).replace(cada_sentenca, f\"<mark>{cada_sentenca}</mark>\")\n else:\n texto_final += str(cada_sentenca)\n texto_final += ' '\n\ndisplay(HTML(f\"\"\"{texto_final}\"\"\"))", "_____no_output_____" ] ], [ [ "# Fim da Tarefa 1", "_____no_output_____" ], [ "## Uso da lib Goose3", "_____no_output_____" ] ], [ [ "from goose3 import Goose", "_____no_output_____" ], [ "g = Goose()\nurl = 'https://www.techtudo.com.br/noticias/2017/08/o-que-e-replika-app-usa-inteligencia-artificial-para-criar-um-clone-seu.ghtml'\nmateria = g.extract(url)", "_____no_output_____" ], [ "materia.title", "_____no_output_____" ], [ "materia.tags", "_____no_output_____" ], [ "materia.infos", "_____no_output_____" ], [ "materia.cleaned_text", "_____no_output_____" ] ], [ [ "# Tarefa 2", "_____no_output_____" ] ], [ [ "frequencia_palavras.keys()", "_____no_output_____" ], [ "frequencia_palavras", "_____no_output_____" ], [ "frase = \"\"\"Algoritmos de aprendizados supervisionados utilizam dados coletados\"\"\".split(' ')\nfrequencia_palavras_frase = []\n\nfor palavra in frase:\n for freq_palavra in frequencia_palavras:\n if palavra in freq_palavra:\n frequencia_palavras_frase.append([palavra, 1])\n\nfrequencia_palavras_frase", "_____no_output_____" ], [ "for x in frequencia_palavras:\n print(x)", "correr\ncaracterístico\ndar\ninteligente\naprendizado\ncoletados\npartir\nestruturar\nestatístico\nalgoritmo\nsupervisionar\nutilizar\nconjuntar\nextrair\npoder\nser\ncorrido\nestabelecer\nrelação\ninteligência\nenquanto\nquantitativo\nmodelo\nmáquina\nconstruir\nreconhecimento\natividades\nhumano\n" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# https://pbpython.com/bullet-graph.html", "_____no_output_____" ] ], [ [ "import matplotlib.pyplot as plt\nimport seaborn as sns\nfrom matplotlib.ticker import FuncFormatter\n\n%matplotlib inline", "_____no_output_____" ], [ "# display a palette of 5 shades of green\nsns.palplot(sns.light_palette(\"green\", 5))", "_____no_output_____" ], [ "# 8 different shades of purple in reverse order\nsns.palplot(sns.light_palette(\"purple\", 8, reverse=True))", "_____no_output_____" ], [ "# define the values we want to plot\nlimits = [80, 100, 150] # 3 ranges: 0-80, 81-100, 101-150\ndata_to_plot = (\"Example 1\", 105, 120) #“Example” line with a value of 105 and target line of 120", "_____no_output_____" ], [ "# blues color palette\npalette = sns.color_palette(\"Blues_r\", len(limits))", "_____no_output_____" ], [ "# build the stacked bar chart of the ranges\nfig, ax = plt.subplots()\nax.set_aspect('equal')\nax.set_yticks([1])\nax.set_yticklabels([data_to_plot[0]])\nax.axvline(data_to_plot[2], color=\"gray\", ymin=0.10, ymax=0.9)\n\nprev_limit = 0\nfor idx, lim in enumerate(limits):\n ax.barh([1], lim-prev_limit, left=prev_limit, height=15, color=palette[idx])\n prev_limit = lim", "_____no_output_____" ], [ "# Draw the value we're measuring\nax.barh([1], data_to_plot[1], color='black', height=5)", "_____no_output_____" ], [ "ax.axvline(data_to_plot[2], color=\"gray\", ymin=0.10, ymax=0.9)\n", "_____no_output_____" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Was Air Quality Affected in Countries or Regions Where COVID-19 was Most Prevalent?\n**By: Arpit Jain, Maria Stella Vardanega, Tingting Cao, Christopher Chang, Mona Ma, Fusu Luo**", "_____no_output_____" ], [ "---", "_____no_output_____" ], [ "## Outline \n\n#### I. Problem Definition & Data Source Description \n\n 1. Project Objectives \n 2. Data Source\n 3. Dataset Preview\n\n#### II. What are the most prevalent pollutants? \n\n#### III. What were the pollutant levels in 2019 and 2020 globally, and their averages?\n\n 1. Selecting Data from 2019 and 2020 with air pollutant information\n 2. Monthly Air Pollutant Data from 2019 \n 3. Monthly Air Pollutant Data from 2020 \n\n#### IV: What cities had the highest changes in pollutant air quality index during COVID-19?\n\n 1. 10 cities with most air quality index reduction for each pollutant \n 2. Cities with more than 50 percent AQI decrease and 50 AQI decrease for each air pollutants\n\n#### V: Regression analysis on COVID-19 cases and pollutant Air Quality Index Globally\n\n#### VI: When were lockdowns implemented for each country?\n\n#### VII: How did Air Quality change in countries with low COVID-19 cases (NZ, AUS, TW) and high COVID-19 cases (US, IT,CN)?\n 1. Countries with high COVID cases\n 2. Countries with low COVID cases\n\n#### VIII: Conclusion\n\n#### IX: Public Tableau Dashboards", "_____no_output_____" ], [ "---", "_____no_output_____" ], [ "## I. Problem Definition & Data Source Description", "_____no_output_____" ], [ "#### 1. Project Objectives", "_____no_output_____" ], [ "Air pollution, as one of the most serious environmental problems confronting our civilization, is the presence of toxic gases and particles in the air at levels that pose adverse effects on global climate and lead to public health risk and disease. Exposure to elevated levels of air pollutants has been implicated in a diverse set of medical conditions including cardiovascular and respiratory mortality, lung cancer and autism. \n\nAir pollutants come from natural sources such as wildfires and volcanoes, as well as are highly related to human activities from mobile sources (such as cars, buses and planes) or stationary sources (such as industrial factories, power plants and wood burning fireplaces). However, in the past year, the COVID-19 pandemic has caused unprecedented changes to the our work, study and daily activities, subsequently led to major reductions in air pollutant emissions. And our team would like take this opportunity to examine the air quality in the past two years and look on how the air quality was impacted in countries and cities where the corona-virus was prevalent.", "_____no_output_____" ], [ "#### 2. Data Source", "_____no_output_____" ], [ "**Data Source Description:** In this project, we downloaded worldwide air quality data for Year 2019 and 2020 from the Air Quality Open Data Platform (https://aqicn.org/data-platform/covid19/), which provides historical air quality index and meteorological data for more than 380 major cities across the world. We used air quality index data in 2019 as baseline to find the air quality changes during COVID in 2020. In addition we joined the data with geographic location information from https://aqicn.org/data-platform/covid19/airquality-covid19-cities.json to get air quality index for each pollutant at city-level. According to the data source provider, the data for each major cities is based on the average (median) of several stations. The data set provides min, max, median and standard deviation for each of the air pollutant species in the form of Air Quality Index (AQI) that are converted from raw concentration based on the US Environmental Protection Agency (EPA) standard.", "_____no_output_____" ], [ "The United States EPA list the following as the criteria at this website (https://www.epa.gov/criteria-air-pollutants/naaqs-table): Carbon Monoxide (CO), Nitrogen Dioxide (NO2), Ozone (O3), Particle Pollution (PM2.5) + (PM10), and finally Sulfur Dioxide (SO2). For the particle pollution the numbers stand for the size of the particles. PM2.5 means particles that are 2.5 micrometers and smaller, while PM10 means particles that are 10 micrometers and smaller. https://www.epa.gov/pm-pollution/particulate-matter-pm-basics. Particle Pollution typically includes Dust, Dirt, and Smoke. Our dataset covers most of the criteria pollutants (PM2.5, PM10, Ozone, SO2, NO2 and CO), and meteorological parameters such as temperature, wind speed, dew point, relative humidity. Air quality index basics are shown in the figure below. ", "_____no_output_____" ], [ "<img src=\"https://github.com/ttcao63/775team_project_b2_t2/blob/main/AQI%20basics.PNG?raw=true\" align=\"center\"/>", "_____no_output_____" ], [ "(source: https://www.airnow.gov/aqi/aqi-basics/)", "_____no_output_____" ], [ "#### 3. Preview of the Dataset", "_____no_output_____" ] ], [ [ "%%bigquery\nSELECT * FROM `ba775-team2-b2.AQICN.air_quality_data` LIMIT 10", "Query complete after 0.00s: 100%|██████████| 1/1 [00:00<00:00, 485.85query/s] \nDownloading: 100%|██████████| 10/10 [00:01<00:00, 6.05rows/s]\n" ] ], [ [ "---\n### II. What are the most prevalent pollutants?\n\nThis question focuses on the prevalence of the pollutants. From the dataset, the prevalence can be defined geographically from the cities and countries that had recorded the parameters detected times. \n\nTo find the prevalence, our team selected the parameters from situations in how many distinct cities and countries detected the parameter appeared. ", "_____no_output_____" ] ], [ [ "%%bigquery\nSELECT \nParameter,COUNT(distinct(City)) AS number_of_city,\nCOUNT(distinct(Country)) AS number_of_country,string_agg(distinct(Country)) AS list_country\nFROM `ba775-team2-b2.AQICN.air_quality_data`\nGROUP BY Parameter\nORDER BY number_of_city DESC", "Query complete after 0.00s: 100%|██████████| 1/1 [00:00<00:00, 918.39query/s] \nDownloading: 100%|██████████| 20/20 [00:01<00:00, 13.28rows/s]\n" ] ], [ [ "From the result, top 6 parameters are meteorological parameters. And the first air pollutants (which can be harmful to the public health and environment) is PM2.5, followed by NO2 and PM10.\n\nPM2.5 has been detected in 548 cities and 92 countries. \nNO2 has been detected in 528 cities and 64 countries. \nPM10 has been detected in 527 cities and 71 countries.\n\nWe conclude PM2.5, NO2 and PM10 are the most prevalent criteria pollutants from the dataset. All of them are considered criteria pollutants set by EPA.", "_____no_output_____" ], [ "---", "_____no_output_____" ], [ "### III. What were the pollutant levels in 2019 and 2020 globally, and their averages? \n\nThe purpose of this question is to determine the air pollutant levels in 2019 and 2020. The air pollutant levels in 2019 serve as a baseline for the air pollutant levels in 2020. In the previous question we observe the distinct parameters that are within the Air Quality Database. Since the meteorological parameters are not needed for the project, we can exclude them, and only focus on the air pollutants. \n\nThe first step is create a table where the parameters are only air pollutants and from the years 2019 and 2020. The next step was to select all the rows from each year, that had a certain parameter, and to union them all. This process was done for all six parameters for both years. ", "_____no_output_____" ], [ "#### 1. Selecting Data from 2019 and 2020 with air pollutant information ", "_____no_output_____" ] ], [ [ "%%bigquery\nSELECT Date, Country, City, lat as Latitude, lon as Longitude, pop as Population, Parameter as Pollutant, median as Pollutant_level\nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE (extract(year from date) = 2019 OR extract(year from date) = 2020) AND parameter IN ('co', 'o3','no2','so2','pm10',\n'pm25')\nORDER BY Country, Date;", "Query complete after 0.00s: 100%|██████████| 2/2 [00:00<00:00, 1175.70query/s] \nDownloading: 100%|██████████| 1968194/1968194 [00:02<00:00, 804214.66rows/s] \n" ] ], [ [ "As we can see after filtering the tables for only the air pollutants we have 1.9 million rows. From here we split the data into 2019 data and 2020 data. ", "_____no_output_____" ], [ "#### 2. Monthly Air Pollutant Data from 2019 ", "_____no_output_____" ] ], [ [ "%%bigquery\nSELECT extract(month from date) Month, Parameter as Pollutant,Round(avg(median),2) as Avg_Pollutant_Level_2019\nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2019 AND parameter IN ('co')\nGROUP BY Month, Parameter\n\nUNION ALL \n\nSELECT extract(month from date) Month, Parameter ,Round(avg(median),2)\nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2019 AND parameter IN ('o3')\nGROUP BY Month, Parameter\n\nUNION ALL \n\nSELECT extract(month from date) Month, Parameter ,Round(avg(median),2) \nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2019 AND parameter IN ('no2')\nGROUP BY Month, Parameter\n\nUNION ALL \n\nSELECT extract(month from date) Month, Parameter ,Round(avg(median),2) \nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2019 AND parameter IN ('so2')\nGROUP BY Month, Parameter\n\nUNION ALL \n\nSELECT extract(month from date) Month, Parameter ,Round(avg(median),2) \nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2019 AND parameter IN ('pm10')\nGROUP BY Month, Parameter\n\nUNION ALL \n\nSELECT extract(month from date) Month, Parameter ,Round(avg(median),2) \nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2019 AND parameter IN ('pm25')\nGROUP BY Month, Parameter\n\n\nORDER BY Month; ", "Query complete after 0.00s: 100%|██████████| 8/8 [00:00<00:00, 4930.85query/s] \nDownloading: 100%|██████████| 72/72 [00:01<00:00, 64.20rows/s] \n" ] ], [ [ "This query represents the average pollutant level for each air pollutant globally for each month. We do this again for the 2020 data. ", "_____no_output_____" ], [ "#### 3. Monthly Air Pollutant Data from 2020 ", "_____no_output_____" ] ], [ [ "%%bigquery\nSELECT extract(month from date) Month, Parameter as Pollutant,Round(avg(median),2) as Avg_Pollutant_Level_2020\nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2020 AND parameter IN ('co')\nGROUP BY Month, Parameter\n\nUNION ALL \n\nSELECT extract(month from date) Month, Parameter ,Round(avg(median),2)\nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2020 AND parameter IN ('o3')\nGROUP BY Month, Parameter\n\nUNION ALL \n\nSELECT extract(month from date) Month, Parameter ,Round(avg(median),2) \nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2020 AND parameter IN ('no2')\nGROUP BY Month, Parameter\n\nUNION ALL \n\nSELECT extract(month from date) Month, Parameter ,Round(avg(median),2) \nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2020 AND parameter IN ('so2')\nGROUP BY Month, Parameter\n\nUNION ALL \n\nSELECT extract(month from date) Month, Parameter ,Round(avg(median),2) \nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2020 AND parameter IN ('pm10')\nGROUP BY Month, Parameter\n\nUNION ALL \n\nSELECT extract(month from date) Month, Parameter ,Round(avg(median),2) \nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE extract(year from date) = 2020 AND parameter IN ('pm25')\nGROUP BY Month, Parameter\n\n\nORDER BY Month; ", "Query complete after 0.00s: 100%|██████████| 8/8 [00:00<00:00, 4959.27query/s] \nDownloading: 100%|██████████| 72/72 [00:01<00:00, 51.38rows/s]\n" ] ], [ [ "When comparing the data there isn't a noticeable difference in global pollutant levels from 2019 to 2020, which leads to the hypothesis of pollutant levels being regional rather than global. This might also mean that whatever effects might be occurring from COVID-19 cases, and lockdowns are short-term enough that the average monthly air pollutant is not capturing small intricacies in the data. We can further narrow down the data by analyzing data from when lockdowns were occurring in different countries, regions, and even cities. ", "_____no_output_____" ], [ "---", "_____no_output_____" ], [ "### IV: What cities had the highest changes in pollutant air quality index during COVID-19? ", "_____no_output_____" ], [ "In this question, we are trying to find cities with most air quality improvement during COVID, and cities with longest time of certain level AQI reduction.", "_____no_output_____" ], [ "#### 1. 10 cities with most air quality index reduction for each pollutant\nMaking queries and creating tables to find monthly average air quality index (AQI) for all pollutants at city level", "_____no_output_____" ], [ "We are using data in 2019 as a baseline and computing AQI differences and percent differences. Negative difference values indicates air quality index decrease, corresponding to an air quality improvement, and positive difference values indicate air quality index increases, corresponding to an air quality deterioration.", "_____no_output_____" ] ], [ [ "%%bigquery \nCREATE OR REPLACE TABLE AQICN.pollutant_diff_daily_aqi_less_than_500\nAS\n(\nSELECT A.Date AS Date_2020,B.Date AS Date_2019,A.Country,A.City,A.lat,A.lon,A.Parameter,A.pop,A.median AS aqi_2020,B.median AS aqi_2019,(A.median-B.median) AS aqi_diff, ROUND((A.median-B.median)/B.median*100,2) AS aqi_percent_diff\nFROM\n(SELECT * FROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE Parameter in ('pm25','pm10','o3','no2','co','so2') AND EXTRACT(Year FROM Date) = 2020 AND median > 0 AND median < 500) AS A\nINNER JOIN\n(SELECT * FROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE Parameter in ('pm25','pm10','o3','no2','co','so2') AND EXTRACT(Year FROM Date) = 2019 AND median > 0 AND median < 500) AS B\nON A.City = B.City\nWHERE EXTRACT(MONTH FROM A.Date) = EXTRACT(MONTH FROM B.Date) AND EXTRACT(DAY FROM A.Date) = EXTRACT(DAY FROM B.Date) AND A.Parameter = B.Parameter\nORDER BY City,Date_2020\n)", "Query complete after 0.00s: 100%|██████████| 4/4 [00:00<00:00, 2424.10query/s] \n" ], [ "%%bigquery \nCREATE OR REPLACE TABLE AQICN.pollutant_diff_monthly_aqi\nAS\nSELECT EXTRACT(month FROM Date_2020) AS month_2020,EXTRACT(month FROM Date_2019) AS month_2019,\n Country,City,lat,lon,Parameter,ROUND(AVG(aqi_2020),1) AS monthly_avg_aqi_2020,\n ROUND(AVG(aqi_2019),1) AS monthly_avg_aqi_2019,(ROUND(AVG(aqi_2020),1)-ROUND(AVG(aqi_2019),1)) AS aqi_diff_monthly,\n ROUND((AVG(aqi_2020)-AVG(aqi_2019))/AVG(aqi_2019)*100,2) AS aqi_percent_diff_monthly\nFROM AQICN.pollutant_diff_daily_aqi_less_than_500\nGROUP BY month_2020,month_2019,Country,City,lat,lon,Parameter\n", "Query complete after 0.00s: 100%|██████████| 4/4 [00:00<00:00, 1818.67query/s] \n" ], [ "%%bigquery\nSELECT *\nFROM AQICN.pollutant_diff_monthly_aqi\nORDER BY Parameter,month_2020,Country\nLIMIT 10", "Query complete after 0.00s: 100%|██████████| 1/1 [00:00<00:00, 368.89query/s] \nDownloading: 100%|██████████| 10/10 [00:01<00:00, 6.39rows/s]\n" ] ], [ [ "Order by monthly average AQI difference to find cities having top 10 air quality index reduction for each pollutant", "_____no_output_____" ] ], [ [ "%%bigquery \n\nCREATE OR REPLACE TABLE AQICN.top_10_cites_most_pollutant_percent_diff_monthly\nAS\n(SELECT * \nFROM AQICN.pollutant_diff_monthly_aqi\nWHERE Parameter = 'co'\nORDER BY aqi_percent_diff_monthly\nLIMIT 10)\nUNION ALL\n(SELECT * \nFROM AQICN.pollutant_diff_monthly_aqi\nWHERE Parameter = 'o3'\nORDER BY aqi_percent_diff_monthly\nLIMIT 10)\nUNION ALL\n(SELECT * \nFROM AQICN.pollutant_diff_monthly_aqi\nWHERE Parameter = 'no2'\nORDER BY aqi_percent_diff_monthly\nLIMIT 10)\nUNION ALL\n(SELECT * \nFROM AQICN.pollutant_diff_monthly_aqi\nWHERE Parameter = 'pm25'\nORDER BY aqi_percent_diff_monthly\nLIMIT 10)\nUNION ALL\n(SELECT * \nFROM AQICN.pollutant_diff_monthly_aqi\nWHERE Parameter = 'pm10'\nORDER BY aqi_percent_diff_monthly\nLIMIT 10)", "Query complete after 0.00s: 100%|██████████| 3/3 [00:00<00:00, 1727.47query/s] \n" ], [ "%%bigquery \nSELECT *\nFROM AQICN.top_10_cites_most_pollutant_percent_diff_monthly\nORDER BY Parameter,aqi_percent_diff_monthly\nLIMIT 10", "Query complete after 0.00s: 100%|██████████| 1/1 [00:00<00:00, 494.55query/s] \nDownloading: 100%|██████████| 10/10 [00:01<00:00, 5.25rows/s]\n" ] ], [ [ "Order by monthly average percent AQI difference to find cities having top 10 most air quality index reduction for each pollutant", "_____no_output_____" ] ], [ [ "%%bigquery \n\nCREATE OR REPLACE TABLE AQICN.top_10_cites_most_pollutant_diff_monthly\nAS\n(SELECT * \nFROM AQICN.pollutant_diff_monthly_aqi\nWHERE Parameter = 'pm25'\nORDER BY aqi_diff_monthly\nLIMIT 10)\nUNION ALL\n(SELECT * \nFROM AQICN.pollutant_diff_monthly_aqi\nWHERE Parameter = 'o3'\nORDER BY aqi_diff_monthly\nLIMIT 10)\nUNION ALL\n(SELECT * \nFROM AQICN.pollutant_diff_monthly_aqi\nWHERE Parameter = 'pm10'\nORDER BY aqi_diff_monthly\nLIMIT 10)\nUNION ALL\n(SELECT * \nFROM AQICN.pollutant_diff_monthly_aqi\nWHERE Parameter = 'no2'\nORDER BY aqi_diff_monthly\nLIMIT 10)\nUNION ALL\n(SELECT * \nFROM AQICN.pollutant_diff_monthly_aqi\nWHERE Parameter = 'so2'\nORDER BY aqi_diff_monthly\nLIMIT 10)\nUNION ALL\n(SELECT * \nFROM AQICN.pollutant_diff_monthly_aqi\nWHERE Parameter = 'co'\nORDER BY aqi_diff_monthly\nLIMIT 10)", "Query complete after 0.00s: 100%|██████████| 3/3 [00:00<00:00, 1556.52query/s] \n" ], [ "%%bigquery \nSELECT *\nFROM AQICN.top_10_cites_most_pollutant_diff_monthly\nORDER BY Parameter,aqi_diff_monthly\nLIMIT 10", "Query complete after 0.00s: 100%|██████████| 1/1 [00:00<00:00, 446.25query/s] \nDownloading: 100%|██████████| 10/10 [00:01<00:00, 6.48rows/s]\n" ] ], [ [ "#### 2. Cities with more than 50 percent AQI decrease and 50 AQI decrease for each air pollutants", "_____no_output_____" ], [ "Reason: the higher the AQI, the unhealthier the air will be, especially for sensitive groups such as people with heart and lung disease, elders and children. A major reduction or percent reduction in AQI for long period of time implies a high air quality impact from the COIVD pandemic.", "_____no_output_____" ] ], [ [ "%%bigquery\n\nSELECT City,Country,Parameter,COUNT(*) AS num_month_mt_50_per_decrease FROM AQICN.pollutant_diff_monthly_aqi\nWHERE aqi_percent_diff_monthly < -50 AND aqi_diff_monthly < -50\nGROUP BY City,Country,Parameter\nORDER BY Parameter,COUNT(*) DESC\nLIMIT 10", "Query complete after 0.00s: 100%|██████████| 1/1 [00:00<00:00, 881.71query/s] \nDownloading: 100%|██████████| 10/10 [00:01<00:00, 6.79rows/s]\n" ] ], [ [ "---", "_____no_output_____" ], [ "Results\n\nDuring the pandemic, cities getting most air qualities improvements in terms of percent AQI differences for each pollutant are:\n\nCO: United States Portland, Chile Talca and Mexico Aguascalientes;\nNO2: Iran Qom, South Africa Middelburg and Philippines Butuan;\nSO2: Greece Athens, Mexico Mérida and Mexico San Luis Potosí;\nOzone: Mexico Aguascalientes, United States Queens and United States The Bronx;\nPM 10: India Gandhinagar, China Hohhot and Israel Tel Aviv;\nPM 2.5: Mexico Mérida, Tajikistan Dushanbe, Bosnia and Herzegovina Sarajevo, Turkey Erzurum, China Qiqihar and India Gandhinagar;\n\n\nCities getting at least 50% and 50 AQI reduction with longest time:\n\nCO: United States Portland, 3 out of 12 months;\nNO2: Iran Qom, 5 out of 12 months;\nO3: Mexico Aguascalientes, 5 out of 12 months;\nPM25: several cities including Iran Kermanshah, Singapore Singapore, AU Sydney and Canberra, 1 out of 12 months;\nPM10: India Gandhinagar and Bhopal, 2 out of 12 months;\nSO2: Mexico Mérida 5 out of 12 months.\n", "_____no_output_____" ], [ "---", "_____no_output_____" ], [ "### V: Regression analysis on COVID-19 cases and pollutant Air Quality Index Globally", "_____no_output_____" ], [ "The purpose of this part is to find the differences in AQI between 2019 and 2020, also the percentage changes for four parameters which include (NO,NO2,PM2.5 and O3), then join with the COVID confirmed table to find the regression between the AQI and the new confirmed case for each air pollutant.", "_____no_output_____" ] ], [ [ "%%bigquery\nselect A.month,A.month_n, A.country,A.parameter,round((B.avg_median_month- A.avg_median_month),2) as diff_avg,\n(B.avg_median_month - A.avg_median_month)/A.avg_median_month as diff_perc\nfrom \n(SELECT FORMAT_DATETIME(\"%B\", date) month,EXTRACT(year FROM date) year, EXTRACT(month FROM date) month_n, country,parameter,round(avg(median),2) as avg_median_month \nFROM `AQICN.Arpit_Cleaned_Data2`\nWHERE Parameter IN ('co','no2','o3','pm25') AND EXTRACT(year FROM date) = 2019\nGROUP by 1,2,3,4,5\nORDER BY country, parameter) A\nleft join \n(SELECT FORMAT_DATETIME(\"%B\", date) month,EXTRACT(year FROM date) year, EXTRACT(month FROM date) month_n, country,parameter,round(avg(median),2) as avg_median_month \nFROM `AQICN.Arpit_Cleaned_Data2`\nWHERE Parameter IN ('co','no2','o3','pm25') AND EXTRACT(year FROM date) = 2020\nGROUP by 1,2,3,4,5\nORDER BY country, parameter) B\nusing (month,country,parameter,month_n)\nwhere A.avg_median_month >0", "Query complete after 0.00s: 100%|██████████| 5/5 [00:00<00:00, 2448.51query/s] \nDownloading: 100%|██████████| 2906/2906 [00:01<00:00, 1788.29rows/s]\n" ], [ "%%bigquery\nselect A.*,confirmed,B.country as country_name\nfrom `all_para_20_19.all_para_20_19_diff` as A\ninner join `covid_population.covid _pop` as B\non A.country = B.country_code2 and A.month = B.month and A.month_n = B.month_n\nwhere B.year = 2020\norder by A.country,A.month_n", "Query complete after 0.00s: 100%|██████████| 3/3 [00:00<00:00, 1451.15query/s] \nDownloading: 100%|██████████| 2789/2789 [00:01<00:00, 1960.44rows/s]\n" ] ], [ [ "Using Bigquery ML to find liner regression between diff_avg for each parameter and confirmed cases\n\n(Example showing below is that parameter = co; x = confirmed; y=diff_avg --AQI changes)", "_____no_output_____" ] ], [ [ "%%bigquery\nCREATE OR REPLACE MODEL `all_para_20_19.all_para_20_19_diff_covid_model`\n# Specify options\nOPTIONS\n (model_type='linear_reg',\n input_label_cols=['diff_avg']) AS\n# Provide training data\nSELECT\nconfirmed,\ndiff_avg\nFROM\n `all_para_20_19.all_para_20_19_diff_covid`\nWHERE\n parameter = 'co'\nand diff_avg is not null\n", "_____no_output_____" ] ], [ [ "Evaluating the model to find out r2_score for each monthly average air pollutant AQI changes vs monthly confirmed new cases linear regression model.\nExample showing below is Evaluation for country level monthly average CO AQI vs monthly new confirmed COVID cases model: ", "_____no_output_____" ] ], [ [ "%%bigquery\nSELECT * FROM\nML.EVALUATE(\n MODEL `all_para_20_19.all_para_20_19_diff_covid_model`, # Model name\n # Table to evaluate against\n (SELECT\nconfirmed,\ndiff_avg\nFROM\n `all_para_20_19.all_para_20_19_diff_covid`\nWHERE\n parameter = 'co'\nand diff_avg is not null\n \n )\n)", "Query complete after 0.00s: 100%|██████████| 3/3 [00:00<00:00, 1508.20query/s] \nDownloading: 100%|██████████| 1/1 [00:01<00:00, 1.86s/rows]\n" ] ], [ [ "Evaluation for country level monthly average PM2.5 AQI changes vs monthly new confirmed COVID cases model: ", "_____no_output_____" ], [ "<img src=\"https://github.com/ttcao63/775team_project_b2_t2/blob/main/pm25_aqi_confirmed_case.png?raw=true\" align=\"center\" width=\"800\"/>", "_____no_output_____" ], [ "Evaluation for country level monthly average NO2 AQI changes vs monthly new confirmed COVID cases model: ", "_____no_output_____" ], [ "<img src=\"https://github.com/ttcao63/775team_project_b2_t2/blob/main/no2_aqi_confirmed_case.png?raw=true\" align=\"center\" width=\"800\"/>", "_____no_output_____" ], [ "Evaluation for country level monthly average O3 AQI changes vs monthly new confirmed COVID cases model: ", "_____no_output_____" ], [ "<img src=\"https://github.com/ttcao63/775team_project_b2_t2/blob/main/o3_aqi_confirmed_case.png?raw=true\" align=\"center\" width=\"800\"/>", "_____no_output_____" ], [ "We have also conducted log transformation of x-variables for linear regression, the most correlated data is PM 2.5 AQI changes vs LOG(confirmed case). Visualization is shown below.", "_____no_output_____" ], [ "<img src=\"https://github.com/ttcao63/775team_project_b2_t2/blob/main/Viz_PM25_Regression.png?raw=true\" align=\"center\" width=\"800\"/>", "_____no_output_____" ], [ "We can see an overall AQI changes from 2019 to 2020. However, after running regression for four air pollutants, model R-squares are less than 0.02, indicating a weak linear relationship between the air quality index changes and the numbers of new confirmed COVID cases. The result makes sense because there are complicated physical and chemical process involved in formation and transportation of air pollution, thus factors such as the weather, energy source, and terrain could also impact the AQI changes. Also, the dramatic increase of new COVID cases might not affect people's response in a way reducing outdoor activities, especially when \"stay at home order\" is partially lifted.\n\nIn this case, we decide to specifically study some countries during their lockdown period and examine the AQI changes.", "_____no_output_____" ], [ "---", "_____no_output_____" ], [ "### VI: When were lockdowns implemented for each country? ", "_____no_output_____" ], [ "Lockdown Dates per Country\n\nChina: Jan 23 - April 8, 2020 (Wuhan 76 day lockdown)\n\nUSA: March 19 - April 7, 2020 \n\nItaly: March 9 - May 18, 2020\n\nTaiwan: No lockdowns in 2020. Lockdown started in July 2021. \n\nAustralia: March 18 - May/June 2020\n\nNew Zealand: March 25 - May/June 2020", "_____no_output_____" ], [ "From the previous regression model we can see that the there was very little correlation between AQI and confirmed cases, and one of the main reasons is that confirmed cases could not accurately capture human activity. To compensate for this, we narrowed down the dates of our pollutant data in order to compare the pollutant levels only during lockdown periods in 2019 and 2020 for the countries where COVID-19 was most prevalent: China, USA, Italy, and those that COVID-19 wasn't as prevalent: Taiwan, Australia, and New Zealand. We came to a conclusion that most lockdown periods started from mid March to April, May, or June, except for China, which started their lockdown late January until April of 2020. To generalize the lockdown dates for countries other than China, the SQL query included dates from the beginning of March to the end of June. As for China, the query included specific dates from January 23 to April 8th of 2020, which is the Wuhan 76 day lockdown. ", "_____no_output_____" ] ], [ [ "%%bigquery\nSELECT country, date, parameter, AVG(count) AS air_quality\nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE date BETWEEN '2020-03-01' AND '2020-06-30'\nAND country in ('US','IT','AU','NZ','TW')\nGROUP BY country, parameter, date\nORDER BY date", "Query complete after 0.00s: 100%|██████████| 1/1 [00:00<00:00, 866.23query/s] \nDownloading: 100%|██████████| 7741/7741 [00:01<00:00, 4628.58rows/s]\n" ], [ "%%bigquery\nSELECT country, date, parameter, AVG(count) AS air_quality\nFROM `ba775-team2-b2.AQICN.air_quality_data`\nWHERE date BETWEEN '2020-01-23' AND '2020-04-08'\nAND country = 'CN'\nGROUP BY country, parameter, date\nORDER BY date", "Query complete after 0.00s: 100%|██████████| 1/1 [00:00<00:00, 1015.32query/s]\nDownloading: 100%|██████████| 900/900 [00:01<00:00, 558.11rows/s]\n" ] ], [ [ "---", "_____no_output_____" ], [ "### VII: How did Air Quality change in countries with low COVID-19 cases (NZ, AUS, TW) and high COVID-19 cases (US, IT,CN)?\nThis question was answered by creating separate tables that encompassed the equivalent lockdown periods per country for 2019. Then, the two tables were joined using the parameter and grouped according to country and parameter to create a subsequent table illustrating the percentage change in average pollution from 2019 to 2020 (during the respective lockdown periods). ", "_____no_output_____" ], [ "#### 1. Countries with high COVID cases", "_____no_output_____" ] ], [ [ "%%bigquery\nCREATE OR REPLACE TABLE AQICN.air_quality2019_Italy AS\n SELECT country, parameter, median AS air_quality2019 FROM `ba775-team2-b2.AQICN.air_quality_data`\n WHERE date BETWEEN '2019-03-09' AND '2019-05-18'\n AND country = 'IT'", "Query complete after 0.01s: 100%|██████████| 3/3 [00:00<00:00, 1165.08query/s] \n" ], [ "%%bigquery\nSELECT a2020.country, a2020.parameter, AVG(a2020.median) AS air_quality2020, AVG(air_quality2019) AS air_quality2019,\n (AVG(a2020.median)-AVG(air_quality2019))/AVG(air_quality2019) AS percentage_change\nFROM `ba775-team2-b2.AQICN.air_quality_data` AS a2020\nLEFT JOIN AQICN.air_quality2019_Italy AS a2019\nUSING(parameter)\nWHERE a2020.date BETWEEN '2020-03-09' AND '2020-05-18'\nAND a2020.country = 'IT'\nAND Parameter in ('pm25','pm10','o3','no2','co','so2')\nGROUP BY a2020.country, a2020.parameter\nORDER BY percentage_change", "Query complete after 0.00s: 100%|██████████| 4/4 [00:00<00:00, 2174.90query/s] \nDownloading: 100%|██████████| 6/6 [00:01<00:00, 3.91rows/s]\n" ] ], [ [ "Here we can see that the only pollutant that decreased during the 2020 lockdown in Italy, compared to the respective time period in 2019, was NO2, which decreased by 35.74%.", "_____no_output_____" ] ], [ [ "%%bigquery\nCREATE OR REPLACE TABLE AQICN.air_quality2019_US AS\n SELECT country, parameter, median AS air_quality2019 FROM `ba775-team2-b2.AQICN.air_quality_data`\n WHERE date BETWEEN '2019-03-19' AND '2019-04-07'\n AND country = 'US'", "Query complete after 0.00s: 100%|██████████| 3/3 [00:00<00:00, 1557.48query/s] \n" ], [ "%%bigquery\nSELECT a2020.country, a2020.parameter, AVG(a2020.median) AS air_quality2020, AVG(air_quality2019) AS air_quality2019,\n (AVG(a2020.median)-AVG(air_quality2019))/AVG(air_quality2019) AS percentage_change\nFROM `ba775-team2-b2.AQICN.air_quality_data` AS a2020\nLEFT JOIN AQICN.air_quality2019_US AS a2019\nUSING(parameter)\nWHERE a2020.date BETWEEN '2020-03-19' AND '2020-04-07'\nAND a2020.country = 'US'\nAND Parameter in ('pm25','pm10','o3','no2','co','so2')\nGROUP BY a2020.country, a2020.parameter\nORDER BY percentage_change", "Query complete after 0.00s: 100%|██████████| 4/4 [00:00<00:00, 2138.31query/s] \nDownloading: 100%|██████████| 6/6 [00:01<00:00, 3.36rows/s]\n" ] ], [ [ "In the United States, all the pollutants decreased in 2020 compared to 2019. The largest changes occurred in O3, NO2 and SO2, which decreased by 36.69%, 30.22%, and 27.10% respectively. This indicates that the lockdowns during the COVID-19 pandemic may have positively affected the emission of pollutants in the United States. ", "_____no_output_____" ] ], [ [ "%%bigquery\nCREATE OR REPLACE TABLE AQICN.air_quality2019_China AS\n SELECT country, parameter, median AS air_quality2019 FROM `ba775-team2-b2.AQICN.air_quality_data`\n WHERE date BETWEEN '2019-01-23' AND '2019-04-08'\n AND country = 'CN'", "Query complete after 0.00s: 100%|██████████| 3/3 [00:00<00:00, 1746.17query/s] \n" ], [ "%%bigquery\nSELECT a2020.country, a2020.parameter, AVG(a2020.median) AS air_quality2020, AVG(air_quality2019) AS air_quality2019,\n (AVG(a2020.median)-AVG(air_quality2019))/AVG(air_quality2019) AS percentage_change\nFROM `ba775-team2-b2.AQICN.air_quality_data` AS a2020\nLEFT JOIN AQICN.air_quality2019_China AS a2019\nUSING(parameter)\nWHERE a2020.date BETWEEN '2020-01-23' AND '2020-04-08'\nAND a2020.country = 'CN'\nAND Parameter in ('pm25','pm10','o3','no2','co','so2')\nGROUP BY a2020.country, a2020.parameter\nORDER BY percentage_change", "Query complete after 0.00s: 100%|██████████| 4/4 [00:00<00:00, 1981.72query/s] \nDownloading: 100%|██████████| 6/6 [00:01<00:00, 4.01rows/s]\n" ] ], [ [ "In China, most pollutants decreased in 2020 compared to the same period in 2019. The largest change was in NO2 which decreased by 30.88% compared to the previous year.", "_____no_output_____" ], [ "#### 2. Countries with low COVID cases", "_____no_output_____" ] ], [ [ "%%bigquery\nCREATE OR REPLACE TABLE AQICN.air_quality2019_Taiwan AS\n SELECT country, parameter, median AS air_quality2019 FROM `ba775-team2-b2.AQICN.air_quality_data`\n WHERE EXTRACT(month FROM date) = 07\n AND EXTRACT(year FROM date) = 2019\n AND country = 'TW'", "Query complete after 0.00s: 100%|██████████| 3/3 [00:00<00:00, 1918.71query/s] \n" ], [ "%%bigquery\nSELECT a2020.country, a2020.parameter, AVG(a2020.median) AS air_quality2020, AVG(air_quality2019) AS air_quality2019,\n (AVG(a2020.median)-AVG(air_quality2019))/AVG(air_quality2019) AS percentage_change\nFROM `ba775-team2-b2.AQICN.air_quality_data` AS a2020\nLEFT JOIN AQICN.air_quality2019_Taiwan AS a2019\nUSING(parameter)\nWHERE EXTRACT(month FROM a2020.date) = 07\nAND EXTRACT(year FROM a2020.date) = 2020\nAND a2020.country = 'TW'\nAND Parameter in ('pm25','pm10','o3','no2','co','so2')\nGROUP BY a2020.country, a2020.parameter\nORDER BY percentage_change", "Query complete after 0.00s: 100%|██████████| 4/4 [00:00<00:00, 1830.37query/s] \nDownloading: 100%|██████████| 6/6 [00:01<00:00, 4.02rows/s]\n" ] ], [ [ "Taiwan, which did not experience lockdowns due to COVID-19, also shows a decrease in all pollutant levels. This contradicts our initially hypothesis that countries who experienced more COVID-19 and therefore more lockdowns would have better air quality. ", "_____no_output_____" ] ], [ [ "%%bigquery\nCREATE OR REPLACE TABLE AQICN.air_quality2019_AUS AS\n SELECT country, parameter, median AS air_quality2019 FROM `ba775-team2-b2.AQICN.air_quality_data`\n WHERE date BETWEEN '2019-03-25' AND '2019-05-31'\n AND country = 'NZ'", "Query complete after 0.00s: 100%|██████████| 3/3 [00:00<00:00, 1679.29query/s] \n" ], [ "%%bigquery\nSELECT a2020.country, a2020.parameter, AVG(a2020.median) AS air_quality2020, AVG(air_quality2019) AS air_quality2019,\n (AVG(a2020.median)-AVG(air_quality2019))/AVG(air_quality2019) AS percentage_change\nFROM `ba775-team2-b2.AQICN.air_quality_data` AS a2020\nLEFT JOIN AQICN.air_quality2019_NZ AS a2019\nUSING(parameter)\nWHERE a2020.date BETWEEN '2020-03-25' AND '2020-05-31'\nAND Parameter in ('pm25','pm10','o3','no2','co','so2')\nAND a2020.country = 'NZ'\nGROUP BY a2020.country, a2020.parameter\nORDER BY percentage_change", "Query complete after 0.00s: 100%|██████████| 4/4 [00:00<00:00, 2199.14query/s] \nDownloading: 100%|██████████| 5/5 [00:01<00:00, 3.27rows/s]\n" ] ], [ [ "New Zealand also shows a decrease in all pollutant levels. Nevertheless, New Zealand did go into lockdown for a period and these numbers may reflect the lessened activity due to COVID-19 during that time compared to the equivalent in 2019. ", "_____no_output_____" ] ], [ [ "%%bigquery\nCREATE OR REPLACE TABLE AQICN.air_quality2019_AUS AS\n SELECT country, parameter, median AS air_quality2019 FROM `ba775-team2-b2.AQICN.air_quality_data`\n WHERE date BETWEEN '2019-03-18' AND '2019-05-31'\n AND country = 'AU'", "Query complete after 0.00s: 100%|██████████| 3/3 [00:00<00:00, 1443.66query/s] \n" ], [ "%%bigquery\nSELECT a2020.country, a2020.parameter, AVG(a2020.median) AS air_quality2020, AVG(air_quality2019) AS air_quality2019,\n (AVG(a2020.median)-AVG(air_quality2019))/AVG(air_quality2019) AS percentage_change\nFROM `ba775-team2-b2.AQICN.air_quality_data` AS a2020\nLEFT JOIN AQICN.air_quality2019_AUS AS a2019\nUSING(parameter)\nWHERE a2020.date BETWEEN '2020-03-18' AND '2020-05-31'\nAND Parameter in ('pm25','pm10','o3','no2','co','so2')\nAND a2020.country = 'AU'\nGROUP BY a2020.country, a2020.parameter\nORDER BY percentage_change", "Query complete after 0.00s: 100%|██████████| 4/4 [00:00<00:00, 2131.79query/s] \nDownloading: 100%|██████████| 6/6 [00:01<00:00, 4.52rows/s]\n" ] ], [ [ "Australia shows decreases in most pollutant parameter levels in 2020 compared to respective periods in 2019. ", "_____no_output_____" ], [ "The fact that all tables illustrate decrease in most pollutant parameter levels, except for Italy, seems to contradict our initial hypothesis. Initially, we hypothesized that in countries where COVID-19 was more prevalent, and therefore where there were more lockdowns and less human activity, there would be better pollutant levels. However, when looking at the results of the analysis, one can see that the extent to which COVID-19 was prevalent does not seem to largely affect the pollutant parameter levels considering that regardless of the country they seem to have decreased in 2020 compared to 2019. This may be due to various governmental and public policies regarding climate change that have pushed countries to improve the air quality as well as the general decrease in human activity worldwide due to the pandemic.", "_____no_output_____" ], [ "---", "_____no_output_____" ], [ "### VIII: Conclusion", "_____no_output_____" ], [ "In this project, we used air quality index (AQI) dataset among 380 major cities across the world from 2019 to 2020 to study criteria air pollutant level changes during the COVID pandemic. According to the result, we conclude that the COVID impacts air quality more at regional-level than at global-level, more in a relative short period of time than in a relative long-term. Even though we don't see a strong relationship between air quality changes versus numbers of confirmed COIVD case, we find that lockdowns during the pandemic do make effects on air pollutant levels in different countries due to reduced outdoor human activities.", "_____no_output_____" ], [ "---", "_____no_output_____" ], [ "### IX: Public Tableau Dashboards\n\nTo interact with our public Tableau dashboards please visit: https://public.tableau.com/app/profile/arpit.jain7335/viz/AnalyzingAirQualityDuringCOVID19Pandemic/AirQualityGlobalLevel?publish=yes", "_____no_output_____" ], [ "<img src=\"https://github.com/arp-jain/BA775-team2-b2/blob/main/Air%20Quality%20Global%20Level.png?raw=true\" align=\"center\"/>", "_____no_output_____" ], [ "<img src=\"https://github.com/arp-jain/BA775-team2-b2/blob/main/Air%20Quality%20City%20Level.png?raw=true\" align=\"center\"/>", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import numpy as np\nimport pandas as pd\nimport seaborn as sns", "_____no_output_____" ] ], [ [ "## Basics\n\n- All values of a categorical valiable are either in `categories` or `np.nan`.\n\n- Order is defined by the order of `categories`, not the lexical order of the values.\n\n- Internally, the data structure consists of a `categories` array and an integer arrays of `codes`, which point to the values in the `categories` array.\n\n- The memory usage of a categorical variable is proportional to the number of categories plus the length of the data, while that for an object dtype is a constant times the length of the data. As the number of categories approaches the length of the data, memory usage approaches that of object type.\n\n- Categories can be useful in the following scenarios:\n\n - To save memory (if number of categories small relative to number of rows)\n \n - If logical order differs from lexical order (e.g. 'small', 'medium', 'large')\n \n - To signal to libraries that column should be treated as a category (e.g. for plotting)", "_____no_output_____" ], [ "## General best practices\n\nBased on [this](https://towardsdatascience.com/staying-sane-while-adopting-pandas-categorical-datatypes-78dbd19dcd8a) useful article.\n\n- Operate on category values rather than column elements. E.g. to rename categories use `df.catvar.cat.rename_rategories(*args, **kwargs)`, if there is no `cat` method available,\nconsider operating on categories directly with `df.catvar.cat.categories`.\n\n- Merging on categories: the two key things to remember are that 1) Pandas treats categorical variables with different categories as different data types, and 2) category merge keys will only be categories in the merged dataframe if they are of the same data types (i.e. have the same categories), otherwise they will be converted back to objects.\n\n- Grouping on categories: remember that by default we group on all categories, not just those present in the data. More often than not, you'll want to use `df.groupby(catvar, observed=True)` to only use categories observed in the data.", "_____no_output_____" ] ], [ [ "titanic = sns.load_dataset(\"titanic\")\ntitanic.head(2)", "_____no_output_____" ] ], [ [ "## Operations I frequently use", "_____no_output_____" ], [ "### Renaming categories", "_____no_output_____" ] ], [ [ "titanic[\"class\"].cat.rename_categories(str.upper)[:2]", "_____no_output_____" ] ], [ [ "### Appending new categories", "_____no_output_____" ] ], [ [ "titanic[\"class\"].cat.add_categories([\"Fourth\"]).cat.categories", "_____no_output_____" ] ], [ [ "### Removing categories", "_____no_output_____" ] ], [ [ "titanic[\"class\"].cat.remove_categories([\"Third\"]).cat.categories", "_____no_output_____" ] ], [ [ "### Remove unused categories", "_____no_output_____" ] ], [ [ "titanic_small = titanic.iloc[:2]\ntitanic_small", "_____no_output_____" ], [ "titanic_small[\"class\"].cat.remove_unused_categories().cat.categories", "_____no_output_____" ] ], [ [ "### Remove and add categories simultaneously", "_____no_output_____" ] ], [ [ "titanic[\"class\"].value_counts(dropna=False)", "_____no_output_____" ], [ "titanic[\"class\"].cat.set_categories([\"First\", \"Third\", \"Fourth\"]).value_counts(\n dropna=False\n)", "_____no_output_____" ] ], [ [ "### Using string and datetime accessors\n\nThis works as expected, and if the number of distinct categories is small relative to the number of rows, then operating on the categories is faster (because under the hood, pandas applies the change to `categories` and constructs a new series (see [here](https://pandas.pydata.org/pandas-docs/stable/user_guide/categorical.html#string-and-datetime-accessors)) so no need to do this manually as I was inclined to).", "_____no_output_____" ] ], [ [ "cat_class = titanic[\"class\"]\n%timeit cat_class.str.contains('d')\n\nstr_class = titanic[\"class\"].astype(\"object\")\n%timeit str_class.str.contains('d')", "149 µs ± 7.84 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n398 µs ± 16.3 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)\n" ] ], [ [ "## Object creation", "_____no_output_____" ], [ "Convert *sex* and *class* to the same categorical type, with categories being the union of all unique values of both columns.", "_____no_output_____" ] ], [ [ "cols = [\"sex\", \"who\"]\nunique_values = np.unique(titanic[cols].to_numpy().ravel())\ncategories = pd.CategoricalDtype(categories=unique_values)\ntitanic[cols] = titanic[cols].astype(categories)\nprint(titanic.sex.cat.categories)\nprint(titanic.who.cat.categories)", "Index(['child', 'female', 'male', 'man', 'woman'], dtype='object')\nIndex(['child', 'female', 'male', 'man', 'woman'], dtype='object')\n" ], [ "# restore sex and who to object types\ntitanic[cols] = titanic[cols].astype(\"object\")", "_____no_output_____" ] ], [ [ "## Custom order ", "_____no_output_____" ] ], [ [ "df = pd.DataFrame({\"quality\": [\"good\", \"excellent\", \"very good\"]})\ndf.sort_values(\"quality\")", "_____no_output_____" ], [ "ordered_quality = pd.CategoricalDtype([\"good\", \"very good\", \"excellent\"], ordered=True)\ndf.quality = df.quality.astype(ordered_quality)\ndf.sort_values(\"quality\")", "_____no_output_____" ] ], [ [ "## Unique values", "_____no_output_____" ] ], [ [ "small_titanic = titanic.iloc[:2]\nsmall_titanic", "_____no_output_____" ] ], [ [ "`Series.unique` returns values in order appearance, and only returns values that are present in the data.", "_____no_output_____" ] ], [ [ "small_titanic[\"class\"].unique()", "_____no_output_____" ] ], [ [ "`Series.cat.categories` returns all category values.", "_____no_output_____" ] ], [ [ "small_titanic[\"class\"].cat.categories", "_____no_output_____" ] ], [ [ "## References\n\n- [Docs](https://pandas.pydata.org/pandas-docs/stable/user_guide/categorical.html#object-creation)\n\n- [Useful Medium article](https://towardsdatascience.com/staying-sane-while-adopting-pandas-categorical-datatypes-78dbd19dcd8a)", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ] ]

[ "BSD-2-Clause" ]

[ "BSD-2-Clause" ]

[ "BSD-2-Clause" ]

[ [ [ "# Machine Learning 1", "_____no_output_____" ], [ "# Some Concepts", "_____no_output_____" ] ], [ [ "Mean Absolute Error (MAE) is the mean of the absolute value of the errors\nMean Squared Error (MSE) is the mean of the squared errors:\nRoot Mean Squared Error (RMSE) is the square root of the mean of the squared errors\n\nComparing these metrics:\n\nMAE is the easiest to understand because it’s the average error.\nMSE is more popular than MAE because MSE “punishes” larger errors, which tends to be useful in the real world.\nRMSE is even more popular than MSE because RMSE is interpretable in the “y” units.", "_____no_output_____" ], [ "# to get the metrics\n\nfrom sklearn import metrics\n\nprint('MAE:', metrics.mean_absolute_error(y_test, y_pred)) \nprint('MSE:', metrics.mean_squared_error(y_test, y_pred)) \nprint('RMSE:', np.sqrt(metrics.mean_squared_error(y_test, y_pred))) \n\n# Different way, just for illustration\ny_pred = linreg.predict(X_test) \n\nfrom sklearn.metrics import mean_squared_error\nMSE = mean_squared_error(y_test, y_pred)\nprint(MSE)\n\n", "_____no_output_____" ] ], [ [ "#### Predictions", "_____no_output_____" ] ], [ [ "# Lets say that the model inputs are\nX = df[['Weight', 'Volume']]\ny = df['CO2']\n\nregr = linear_model.LinearRegression()\nregr.fit(X, y)\n\n# Simply do that for predicting the CO2 emission of a car where the weight is 2300kg, and the volume is 1300ccm:\npredictedCO2 = regr.predict([[2300, 1300]])\n\nprint(predictedCO2)\n", "_____no_output_____" ] ], [ [ "#### OLS Regression", "_____no_output_____" ] ], [ [ "https://docs.w3cub.com/statsmodels/generated/statsmodels.regression.linear_model.ols.fit_regularized/", "_____no_output_____" ], [ "est=sm.OLS(y, X)\nest = est.fit()\nest.summary()", "_____no_output_____" ] ], [ [ "#### Plotting Errors", "_____no_output_____" ] ], [ [ "# provided that y_test and y_pred have been called (example below)\n# X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)\n# y_pred = linreg.predict(X_test)\n\nsns.distplot((y_test-y_pred),bins=50) ", "_____no_output_____" ] ], [ [ "# Interpretation of Outputs", "_____no_output_____" ], [ "## Multiple Linear Regression", "_____no_output_____" ] ], [ [ "Almost all the real-world problems that you are going to encounter will have more than two variables. \nLinear regression involving multiple variables is called “multiple linear regression” or multivariate linear regression. \nThe steps to perform multiple linear regression are almost similar to that of simple linear regression. \n\nThe difference lies in the evaluation. \nYou can use it to find out which factor has the highest impact on the predicted output and how different variables relate to each other.", "_____no_output_____" ] ], [ [ "# Logistic Regression", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Tensorflow Timeline Analysis on Model Zoo Benchmark between Intel optimized and stock Tensorflow\n\nThis jupyter notebook will help you evaluate performance benefits from Intel-optimized Tensorflow on the level of Tensorflow operations via several pre-trained models from Intel Model Zoo. The notebook will show users a bar chart like the picture below for the Tensorflow operation level performance comparison. The red horizontal line represents the performance of Tensorflow operations from Stock Tensorflow, and the blue bars represent the speedup of Intel Tensorflow operations. The operations marked as \"mkl-True\" are accelerated by MKL-DNN a.k.a oneDNN, and users should be able to see a good speedup for those operations accelerated by MKL-DNN. \n> NOTE : Users need to get Tensorflow timeline json files from other Jupyter notebooks like benchmark_perf_comparison\n first to proceed this Jupyter notebook.", "_____no_output_____" ], [ "<img src=\"images\\compared_tf_op_duration_ratio_bar.png\" width=\"700\">", "_____no_output_____" ], [ "The notebook will also show users two pie charts like the picture below for elapsed time percentage among different Tensorflow operations. \nUsers can easily find the Tensorflow operation hotspots in these pie charts among Stock and Intel Tensorflow.", "_____no_output_____" ], [ "<img src=\"images\\compared_tf_op_duration_pie.png\" width=\"700\">", "_____no_output_____" ], [ "# Get Platform Information ", "_____no_output_____" ] ], [ [ "from profiling.profile_utils import PlatformUtils\nplat_utils = PlatformUtils()\nplat_utils.dump_platform_info()", "_____no_output_____" ] ], [ [ "# Section 1: TensorFlow Timeline Analysis\n## Prerequisites", "_____no_output_____" ] ], [ [ "!pip install cxxfilt\n\n%matplotlib inline\nimport matplotlib.pyplot as plt\nimport tensorflow as tf", "_____no_output_____" ], [ "import pandas as pd\npd.set_option('display.max_rows', 500)\npd.set_option('display.max_columns', 500)\npd.set_option('display.width', 1500)", "_____no_output_____" ] ], [ [ "## List out the Timeline folders", "_____no_output_____" ], [ "First, list out all Timeline folders from previous runs.", "_____no_output_____" ] ], [ [ "import os\nfilenames= os.listdir (\".\") \nresult = []\nkeyword = \"Timeline\"\nfor filename in filenames: \n if os.path.isdir(os.path.join(os.path.abspath(\".\"), filename)): \n if filename.find(keyword) != -1:\n result.append(filename)\nresult.sort()\n\nindex =0 \nfor folder in result:\n print(\" %d : %s \" %(index, folder))\n index+=1", "_____no_output_____" ] ], [ [ "## Select a Timeline folder from previous runs\n#### ACTION: Please select one Timeline folder and change FdIndex accordingly", "_____no_output_____" ] ], [ [ "FdIndex = 3", "_____no_output_____" ] ], [ [ "List out all Timeline json files inside Timeline folder.", "_____no_output_____" ] ], [ [ "import os\nTimelineFd = result[FdIndex]\nprint(TimelineFd)\ndatafiles = [TimelineFd +os.sep+ x for x in os.listdir(TimelineFd) if '.json' == x[-5:]]\nprint(datafiles)\nif len(datafiles) is 0:\n print(\"ERROR! No json file in the selected folder. Please select other folder.\")\nelif len(datafiles) is 1:\n print(\"WARNING! There is only 1 json file in the selected folder. Please select other folder to proceed Section 1.2.\")", "_____no_output_____" ] ], [ [ "> **Users can bypass below Section 1.1 and analyze performance among Stock and Intel TF by clicking the link : [Section 1_2](#section_1_2).**", "_____no_output_____" ], [ "<a id='section_1_1'></a>\n## Section 1.1: Performance Analysis for one TF Timeline result\n### Step 1: Pick one of the Timeline files\n#### List out all the Timeline files first\n", "_____no_output_____" ] ], [ [ "index = 0\nfor file in datafiles:\n print(\" %d : %s \" %(index, file))\n index+=1", "_____no_output_____" ] ], [ [ "#### ACTION: Please select one timeline json file and change file_index accordingly", "_____no_output_____" ] ], [ [ "## USER INPUT\nfile_index=0\n\nfn = datafiles[file_index]\ntfile_prefix = fn.split('_')[0]\ntfile_postfix = fn.strip(tfile_prefix)[1:]\nfn", "_____no_output_____" ] ], [ [ "### Step 2: Parse timeline into pandas format", "_____no_output_____" ] ], [ [ "from profiling.profile_utils import TFTimelinePresenter\ntfp = TFTimelinePresenter(True)\ntimeline_pd = tfp.postprocess_timeline(tfp.read_timeline(fn))\ntimeline_pd = timeline_pd[timeline_pd['ph'] == 'X']", "_____no_output_____" ] ], [ [ "### Step 3: Sum up the elapsed time of each TF operation", "_____no_output_____" ] ], [ [ "tfp.get_tf_ops_time(timeline_pd,fn,tfile_prefix)", "_____no_output_____" ] ], [ [ "### Step 4: Draw a bar chart for elapsed time of TF ops ", "_____no_output_____" ] ], [ [ "filename= tfile_prefix +'_tf_op_duration_bar.png'\ntitle_=tfile_prefix +'TF : op duration bar chart'\nax=tfp.summarize_barh(timeline_pd, 'arg_op', title=title_, topk=50, logx=True, figsize=(10,10))\ntfp.show(ax,'bar')", "_____no_output_____" ] ], [ [ "### Step 5: Draw a pie chart for total time percentage of TF ops ", "_____no_output_____" ] ], [ [ "filename= tfile_prefix +'_tf_op_duration_pie.png'\ntitle_=tfile_prefix +'TF : op duration pie chart'\ntimeline_pd_known = timeline_pd[ ~timeline_pd['arg_op'].str.contains('unknown') ]\nax=tfp.summarize_pie(timeline_pd_known, 'arg_op', title=title_, topk=50, logx=True, figsize=(10,10))\ntfp.show(ax,'pie')\nax.figure.savefig(filename,bbox_inches='tight')", "_____no_output_____" ] ], [ [ "<a id='section_1_2'></a>\n## Section 1.2: Analyze TF Timeline results between Stock and Intel Tensorflow\n### Speedup from MKL-DNN among different TF operations", "_____no_output_____" ], [ "### Step 1: Select one Intel and one Stock TF timeline files for analysis\n#### List out all timeline files in the selected folder", "_____no_output_____" ] ], [ [ "if len(datafiles) is 1:\n print(\"ERROR! There is only 1 json file in the selected folder.\")\n print(\"Please select other Timeline folder from beginnning to proceed Section 1.2.\")\n\nfor i in range(len(datafiles)):\n print(\" %d : %s \" %(i, datafiles[i]))", "_____no_output_____" ] ], [ [ "#### ACTION: Please select one timeline file as a perfomance baseline and the other as a comparison target\nput the related index for your selected timeline file.\nIn general, please put stock_timeline_xxxxx as the baseline.", "_____no_output_____" ] ], [ [ "# perfomance baseline \nBaseline_Index=1\n# comparison target\nComparison_Index=0", "_____no_output_____" ] ], [ [ "#### List out two selected timeline files", "_____no_output_____" ] ], [ [ "selected_datafiles = []\nselected_datafiles.append(datafiles[Baseline_Index])\nselected_datafiles.append(datafiles[Comparison_Index])\nprint(selected_datafiles)", "_____no_output_____" ] ], [ [ "### Step 2: Parsing timeline results into CSV files", "_____no_output_____" ] ], [ [ "%matplotlib agg\nfrom profiling.profile_utils import TFTimelinePresenter\ncsvfiles=[]\n\ntfp = TFTimelinePresenter(True)\nfor fn in selected_datafiles:\n if fn.find('/'):\n fn_nofd=fn.split('/')[1]\n else:\n fn_nofd=fn\n tfile_name= fn_nofd.split('.')[0]\n tfile_prefix = fn_nofd.split('_')[0]\n tfile_postfix = fn_nofd.strip(tfile_prefix)[1:]\n csvpath = TimelineFd +os.sep+tfile_name+'.csv'\n print(csvpath)\n csvfiles.append(csvpath)\n timeline_pd = tfp.postprocess_timeline(tfp.read_timeline(fn))\n timeline_pd = timeline_pd[timeline_pd['ph'] == 'X']\n tfp.get_tf_ops_time(timeline_pd,fn,tfile_prefix)", "_____no_output_____" ] ], [ [ "### Step 3: Pre-processing for the two CSV files", "_____no_output_____" ] ], [ [ "import os\nimport pandas as pd\n\ncsvarray=[]\nfor csvf in csvfiles:\n print(\"read into pandas :\",csvf)\n a = pd.read_csv(csvf)\n csvarray.append(a)\n\na = csvarray[0]\nb = csvarray[1]", "_____no_output_____" ] ], [ [ "### Step 4: Merge two CSV files and caculate the speedup accordingly", "_____no_output_____" ] ], [ [ "import os\nimport pandas as pd\nfdir='merged'\nif not os.path.exists(fdir):\n os.mkdir(fdir)\n \nfpath=fdir+os.sep+'merged.csv'\nmerged=tfp.merge_two_csv_files(fpath,a,b)\nmerged", "_____no_output_____" ] ], [ [ "### Step 5: Draw a bar chart for elapsed time of TF ops among stock TF and Intel TF", "_____no_output_____" ] ], [ [ "%matplotlib inline\nprint(fpath)\ntfp.plot_compare_bar_charts(fpath)\ntfp.plot_compare_ratio_bar_charts(fpath, tags=['','oneDNN ops'])", "_____no_output_____" ] ], [ [ "### Step 6: Draw pie charts for elapsed time of TF ops among stock TF and Intel TF", "_____no_output_____" ] ], [ [ "tfp.plot_compare_pie_charts(fpath)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "diagrams.generic.network.Firewall\ndiagrams.generic.network.Router\ndiagrams.generic.network.Subnet\ndiagrams.generic.network.Switch\ndiagrams.generic.network.VPN\n\ndiagrams.generic.virtualization.Virtualbox\n\ndiagrams.generic.os.Windows", "_____no_output_____" ] ], [ [ "from diagrams import Cluster, Diagram\nfrom diagrams.generic.network import Firewall\nfrom diagrams.generic.network import Router\nfrom diagrams.generic.network import Subnet\nfrom diagrams.generic.network import Switch\nfrom diagrams.generic.virtualization import Virtualbox\nfrom diagrams.generic.os import Windows", "_____no_output_____" ], [ "graph_attr = {\n \"fontsize\": \"28\",\n \"bgcolor\": \"grey\"\n}\n\nwith Diagram(\"My Network Automation with Python Lab\", filename=\"MyLab\", outformat=\"jpg\", graph_attr=graph_attr, show=True):\n my_computer = Windows(\"My Computer\")\n my_home_subnet = Subnet(\"My Home Subnet\")\n \n \n with Cluster(\"Virtualbox\"):\n lab_devs = [Switch(\"sw01\"),\n Switch(\"sw02\"),\n Switch(\"sw03 optional\")]\n\n my_computer >> my_home_subnet >> lab_devs", "_____no_output_____" ], [ "ls", "MyLab.jpg Untitled.ipynb my_lab.png\r\n" ], [ "rm my_lab.png", "_____no_output_____" ], [ "ls", "MyLab.jpg Untitled.ipynb\r\n" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import torchvision, torch\nfrom PIL import Image\nfrom IPython.display import display\nfrom torch import nn\nfrom model import *\nfrom torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential", "_____no_output_____" ], [ "image_path = \"cat.jpg\"\nimage = Image.open(image_path)\n\ntransform = torchvision.transforms.Compose([torchvision.transforms.Resize((32,32)),torchvision.transforms.ToTensor()])\nimage = transform(image)\n\ndisplay(torchvision.transforms.ToPILImage()(image))", "_____no_output_____" ], [ "CIFAR10Model = torch.load(\"CIFAR10Model_trained.pth\")\nprint(CIFAR10Model)", "CIFAR10model(\n (model1): Sequential(\n (0): Conv2d(3, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n (1): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (2): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (4): Conv2d(32, 64, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n (5): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (6): Flatten(start_dim=1, end_dim=-1)\n (7): Linear(in_features=1024, out_features=64, bias=True)\n (8): Linear(in_features=64, out_features=10, bias=True)\n )\n)\n" ], [ "image = torch.reshape(image, (1,3,32,32))\noutput = CIFAR10Model(image)\nprint(output.argmax(1))\nClassSet= ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']", "tensor([3])\n" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "The data is from a number of patients. The 12 first columns (age, an, ..., time) are features that should be used to predict the outcome in the last column (DEATH_EVENT).\n", "_____no_output_____" ] ], [ [ "# Loading some functionality you might find useful. You might want other than this...\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport seaborn as sns\nfrom sklearn import metrics\nfrom sklearn.model_selection import train_test_split, cross_val_score, GridSearchCV\nfrom pandas.plotting import scatter_matrix\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.neighbors import KNeighborsClassifier\nfrom sklearn.tree import DecisionTreeClassifier, plot_tree\nfrom sklearn.ensemble import RandomForestClassifier\nfrom sklearn.svm import SVC\nfrom sklearn.decomposition import PCA\nfrom sklearn.cluster import KMeans\n\n", "_____no_output_____" ], [ "# Downloading data \nurl = 'https://raw.githubusercontent.com/BoBernhardsson/frtn65_exam2022/main/data.csv'\ndata = pd.read_csv(url)\ndata.head()", "_____no_output_____" ], [ "# Picking out features and labels\nX = data.iloc[:,:-1].values\ny = data.iloc[:,-1].values\n(X.shape,y.shape)", "_____no_output_____" ], [ "# select which features to use\n\nX = data.drop(columns=['DEATH_EVENT'])\ny = data.loc[:,'DEATH_EVENT'].values", "_____no_output_____" ], [ "# Creating some initial KNN models and evaluating accuracy \nNdata = data.shape[0]\nfor nr in range(1,10):\n knnmodel = KNeighborsClassifier(n_neighbors = nr)\n knnmodel.fit(X=X,y=y)\n predictions = knnmodel.predict(X=X)\n print('neighbors = {0}: accuracy = {1:.3f}'.format(nr,1-sum(abs(predictions-y))/Ndata))", "neighbors = 1: accuracy = 1.000\nneighbors = 2: accuracy = 0.779\nneighbors = 3: accuracy = 0.779\nneighbors = 4: accuracy = 0.726\nneighbors = 5: accuracy = 0.749\nneighbors = 6: accuracy = 0.719\nneighbors = 7: accuracy = 0.722\nneighbors = 8: accuracy = 0.702\nneighbors = 9: accuracy = 0.709\n" ], [ "features = ['age', 'cr', 'ej', 'pl', 'se1', 'se2', 'time']\nfeatures_categ = ['an','di', 'hi', 'sex', 'sm']\n\n#scale the dataset\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.preprocessing import OneHotEncoder\nfrom sklearn.compose import ColumnTransformer\n\nnumerical_preprocessor = StandardScaler()\n\n#do one-hot encoding for categorical features\ncategorical_preprocessor = OneHotEncoder(handle_unknown=\"ignore\")\npreprocessor = ColumnTransformer([\n ('one-hot-encoder', categorical_preprocessor, features_categ),\n ('standard-scaler', numerical_preprocessor, features)])", "_____no_output_____" ], [ "from sklearn.pipeline import Pipeline\n\n#try Random Forest\nclf = Pipeline(steps = [('preprocessor', preprocessor), ('classifier', RandomForestClassifier(n_estimators=100,max_depth=3))])", "_____no_output_____" ], [ "#split the dataset into trainig and testing\nx_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)", "_____no_output_____" ], [ "clf.fit(x_train, np.ravel(y_train))\nprint(clf.score(x_train,np.ravel(y_train)))\nprint('model score: %.3f' % clf.score(x_test,np.ravel(y_test)))", "0.9121338912133892\nmodel score: 0.733\n" ], [ "#evaluate the model using cross-validation\nscore = cross_val_score(clf, X, np.ravel(y), cv=25)\nprint(\"%0.2f accuracy with a standard deviation of %0.2f\" % (score.mean()*100, score.std()))\nscore_test = clf.score(x_test, np.ravel(y_test))\nprint('Test score: ', '{0:.4f}'.format(score_test*100))", "82.58 accuracy with a standard deviation of 0.15\nTest score: 73.3333\n" ], [ "clf = RandomForestClassifier()\n#find the best model using GridSearch\nparam_grid = {\n 'n_estimators': [100, 1000],\n 'max_depth': [3, 4, 5],\n }\n\nsearch = GridSearchCV(clf, param_grid, cv=4, verbose=1,n_jobs=-1)\n\nsearch.fit(x_train, np.ravel(y_train))\nscore = search.score(x_test, np.ravel(y_test))\nprint(\"Best CV score: {} using {}\".format(search.best_score_, search.best_params_))\nprint(\"Test accuracy: {}\".format(score))", "Fitting 4 folds for each of 6 candidates, totalling 24 fits\nBest CV score: 0.8703389830508474 using {'max_depth': 4, 'n_estimators': 100}\nTest accuracy: 0.75\n" ], [ "randomForestModel = RandomForestClassifier(n_estimators=100,max_depth=5)\n\n#evaluate using cross-validation\nscore=cross_val_score(randomForestModel, X, y, cv=20)", "_____no_output_____" ], [ "randomForestModel.fit(x_train,np.ravel(y_train))\nprint('Training score: ', randomForestModel.score(x_train,np.ravel(y_train)))\nprint('Test score: ', randomForestModel.score(x_test,np.ravel(y_test)))", "Training score: 0.9623430962343096\nTest score: 0.7666666666666667\n" ], [ "#make a prediction and evaluate the performance\ny_pred = randomForestModel.predict(x_test)\n\nscore_new = randomForestModel.score(x_test, y_test)\nprint('Test score: ', score_new)", "Test score: 0.7666666666666667\n" ], [ "import seaborn as sns\nfrom sklearn import metrics\n#confusion matrix\ncm = metrics.confusion_matrix(y_test, y_pred)\nplt.figure(figsize=(10,10))\nsns.heatmap(cm, annot=True, fmt=\".0f\", linewidths=1, square = True);\nplt.ylabel('Actual label');\nplt.xlabel('Predicted label');\nplt.title('Accuracy Score: {0}'.format(score.mean()), size = 15);", "_____no_output_____" ], [ "from sklearn import metrics\n#AUC\nmetrics.plot_roc_curve(randomForestModel, x_test, y_test)", "/usr/local/lib/python3.7/dist-packages/sklearn/utils/deprecation.py:87: FutureWarning: Function plot_roc_curve is deprecated; Function :func:`plot_roc_curve` is deprecated in 1.0 and will be removed in 1.2. Use one of the class methods: :meth:`sklearn.metric.RocCurveDisplay.from_predictions` or :meth:`sklearn.metric.RocCurveDisplay.from_estimator`.\n warnings.warn(msg, category=FutureWarning)\n" ], [ "from sklearn.metrics import classification_report\n\nprint(classification_report(y_test, randomForestModel.predict(x_test)))", " precision recall f1-score support\n\n 0 0.72 0.97 0.83 35\n 1 0.92 0.48 0.63 25\n\n accuracy 0.77 60\n macro avg 0.82 0.73 0.73 60\nweighted avg 0.81 0.77 0.75 60\n\n" ], [ "from pandas import DataFrame\nfeature_df = DataFrame(data.columns.delete(0))\nfeature_df.columns = ['Features']\nfeature_df[\"Feature Importance\"] = pd.Series(randomForestModel.feature_importances_)\n\n#view feature importance according to Random Forest model\nfeature_df", "_____no_output_____" ], [ "#KNN model\n\nclf = Pipeline(steps = [('preprocessor', preprocessor), ('classifier', KNeighborsClassifier(n_neighbors=3))])", "_____no_output_____" ], [ "clf.fit(x_train,np.ravel(y_train))", "_____no_output_____" ], [ "#evaluate the model using cross-validation\nscore = cross_val_score(clf, X, np.ravel(y), cv=25)\nprint(\"%0.2f accuracy with a standard deviation of %0.2f\" % (scores.mean()*100, scores.std()))\nscore_test = clf.score(x_test, np.ravel(y_test))\nprint('Test score: ', '{0:.4f}'.format(score_test*100))", "71.58 accuracy with a standard deviation of 0.08\nTest score: 75.0000\n" ], [ "#make a prediction and evaluate the performance\ny_pred = clf.predict(x_test)\n\nscore_new = clf.score(x_test, y_test)\nprint('Test score: ', score_new)", "Test score: 0.75\n" ], [ "import seaborn as sns\nfrom sklearn import metrics\n\n#confusion matrix\ncm = metrics.confusion_matrix(y_test, y_pred)\nplt.figure(figsize=(10,10))\nsns.heatmap(cm, annot=True, fmt=\".0f\", linewidths=1, square = True);\nplt.ylabel('Actual label');\nplt.xlabel('Predicted label');\nplt.title('Accuracy Score: {0}'.format(score.mean()), size = 15);", "_____no_output_____" ], [ "from sklearn import metrics\n#AUC\nmetrics.plot_roc_curve(clf, x_test, y_test)", "/usr/local/lib/python3.7/dist-packages/sklearn/utils/deprecation.py:87: FutureWarning: Function plot_roc_curve is deprecated; Function :func:`plot_roc_curve` is deprecated in 1.0 and will be removed in 1.2. Use one of the class methods: :meth:`sklearn.metric.RocCurveDisplay.from_predictions` or :meth:`sklearn.metric.RocCurveDisplay.from_estimator`.\n warnings.warn(msg, category=FutureWarning)\n" ], [ "from sklearn.metrics import classification_report\nprint(classification_report(y_test, clf.predict(x_test)))", " precision recall f1-score support\n\n 0 0.70 1.00 0.82 35\n 1 1.00 0.40 0.57 25\n\n accuracy 0.75 60\n macro avg 0.85 0.70 0.70 60\nweighted avg 0.82 0.75 0.72 60\n\n" ], [ "#Boosting model\n\nfrom sklearn.ensemble import GradientBoostingClassifier\n#find the best learning rate\nlearning_rates = [0.05, 0.1, 0.25, 0.5, 0.75, 1]\nfor learning_rate in learning_rates:\n gb = Pipeline(steps = [('preprocessor', preprocessor), ('Classifier', GradientBoostingClassifier(n_estimators=30, learning_rate = learning_rate, max_features=13, max_depth = 3, random_state = 0))])\n gb.fit(x_train, y_train)\n print(\"Learning rate: \", learning_rate)\n print(\"Accuracy score (training): {0:.3f}\".format(gb.score(x_train, y_train)))\n print(\"Accuracy score (validation): {0:.3f}\".format(gb.score(x_test, y_test)))", "Learning rate: 0.05\nAccuracy score (training): 0.946\nAccuracy score (validation): 0.733\nLearning rate: 0.1\nAccuracy score (training): 0.950\nAccuracy score (validation): 0.750\nLearning rate: 0.25\nAccuracy score (training): 0.996\nAccuracy score (validation): 0.800\nLearning rate: 0.5\nAccuracy score (training): 1.000\nAccuracy score (validation): 0.767\nLearning rate: 0.75\nAccuracy score (training): 1.000\nAccuracy score (validation): 0.683\nLearning rate: 1\nAccuracy score (training): 1.000\nAccuracy score (validation): 0.750\n" ], [ "clf = Pipeline(steps = [('preprocessor', preprocessor), ('Classifier', GradientBoostingClassifier(n_estimators= 30, learning_rate = 0.25, max_features=13, max_depth = 3, random_state = 0))])", "_____no_output_____" ], [ "clf.fit(x_train,np.ravel(y_train))", "_____no_output_____" ], [ "#evaluate the models using cross-validation\n\nfrom sklearn.model_selection import cross_val_score\nscores = cross_val_score(clf, X, np.ravel(y), cv=25)\nprint(\"%0.2f accuracy with a standard deviation of %0.2f\" % (scores.mean()*100, scores.std()))\nscore = clf.score(x_test, np.ravel(y_test))\nprint('Test score (Validation): ', '{0:.4f}'.format(score*100))", "79.91 accuracy with a standard deviation of 0.14\nTest score (Validation): 80.0000\n" ], [ "#make a prediction and evaluate the performance\ny_pred = clf.predict(x_test)\n\nscore_test = clf.score(x_test, y_test)\nprint('Test score: ', score_test )", "Test score: 0.8\n" ], [ "import seaborn as sns\nfrom sklearn import metrics\n\n#confusion matrix\ncm = metrics.confusion_matrix(y_test, y_pred)\nplt.figure(figsize=(10,10))\nsns.heatmap(cm, annot=True, fmt=\".0f\", linewidths=1, square = True);\nplt.ylabel('Actual label');\nplt.xlabel('Predicted label');\nplt.title('Accuracy Score: {0}'.format(score.mean()), size = 15);", "_____no_output_____" ], [ "from sklearn import metrics\n#AUC\nmetrics.plot_roc_curve(clf, X, y)", "/usr/local/lib/python3.7/dist-packages/sklearn/utils/deprecation.py:87: FutureWarning: Function plot_roc_curve is deprecated; Function :func:`plot_roc_curve` is deprecated in 1.0 and will be removed in 1.2. Use one of the class methods: :meth:`sklearn.metric.RocCurveDisplay.from_predictions` or :meth:`sklearn.metric.RocCurveDisplay.from_estimator`.\n warnings.warn(msg, category=FutureWarning)\n" ], [ "from sklearn.metrics import classification_report\nprint(classification_report(y_test, clf.predict(x_test)))\n", " precision recall f1-score support\n\n 0 0.77 0.94 0.85 35\n 1 0.88 0.60 0.71 25\n\n accuracy 0.80 60\n macro avg 0.82 0.77 0.78 60\nweighted avg 0.82 0.80 0.79 60\n\n" ] ], [ [ "## Death vs time\nThe boxplot below illustrates the relationship between death and how long time it was between the measurements were taken and the followup event, when the patient health was checked (female=blue, male=orange).\n\nIt is noted that short followup time is highly related to high probability of death, for both sexes. An explanation could be that severly unhealthy patients were followed up earlier, based on medical expert decisions.", "_____no_output_____" ] ], [ [ "fig, ax = plt.subplots(figsize = (8, 8))\nsurvive = data.loc[(data.DEATH_EVENT == 0)].time\ndeath = data.loc[(data.DEATH_EVENT == 1)].time\nprint('time_survived = {:.1f}'.format(survive.mean()))\nprint('time_dead = {:.1f}'.format(death.mean()))\n\nsns.boxplot(data = data, x = 'DEATH_EVENT', y = 'time', hue = 'sex', width = 0.4, ax = ax, fliersize = 3, palette=sns.color_palette(\"pastel\"))\nsns.stripplot(data = data, x = 'DEATH_EVENT', y = 'time', hue = 'sex', size = 3, palette=sns.color_palette())\nax.set(xlabel = 'DEATH', ylabel = \"time [days] \", title = 'The relationship between death and time')\nplt.show()", "time_survived = 158.3\ntime_dead = 70.9\n" ], [ "# If we want to drop time as feature, we can use\nXnew = data.iloc[:,:-2].values", "_____no_output_____" ], [ "# select which features to use\n\nXnew = data.drop(columns=['DEATH_EVENT', 'time', 'sm', 'ej','age','cr','pl','se1','an','di','sex'])\ny = data.loc[:,'DEATH_EVENT'].values", "_____no_output_____" ], [ "features = ['se2','hi']\n#features_categ = ['an','di', 'hi', 'sex']\n\n#scale the dataset\nfrom sklearn.preprocessing import StandardScaler\nfrom sklearn.preprocessing import OneHotEncoder\nfrom sklearn.compose import ColumnTransformer\n\nnumerical_preprocessor = StandardScaler()\n\n#do one-hot encoding for categorical features\ncategorical_preprocessor = OneHotEncoder(handle_unknown=\"ignore\")\npreprocessor = ColumnTransformer([\n #('one-hot-encoder', categorical_preprocessor, features_categ),\n ('standard-scaler', numerical_preprocessor, features)])", "_____no_output_____" ], [ "from sklearn.pipeline import Pipeline\n\n#try Random Forest\nclf = Pipeline(steps = [('preprocessor', preprocessor), ('classifier', RandomForestClassifier(n_estimators=1000,max_depth=3))])", "_____no_output_____" ], [ "#split the dataset into trainig and testing\nx_train_new, x_test_new, y_train_new, y_test_new = train_test_split(Xnew, y, test_size=0.2, random_state=42)", "_____no_output_____" ], [ "clf.fit(x_train_new, np.ravel(y_train_new))\nprint(clf.score(x_train_new,np.ravel(y_train_new)))\nprint('model score: %.3f' % clf.score(x_test_new,np.ravel(y_test_new)))", "0.7280334728033473\nmodel score: 0.617\n" ], [ "clf = RandomForestClassifier()\n\nparam_grid = {\n 'n_estimators': [100, 1000],\n 'max_depth': [3, 4, 5],\n }\n\nsearch = GridSearchCV(clf, param_grid, cv=4, verbose=1,n_jobs=-1)\n\nsearch.fit(x_train_new, np.ravel(y_train_new))\nscore = search.score(x_test_new, np.ravel(y_test_new))\nprint(\"Best CV score: {} using {}\".format(search.best_score_, search.best_params_))\nprint(\"Test accuracy: {}\".format(score))", "Fitting 4 folds for each of 6 candidates, totalling 24 fits\nBest CV score: 0.6903954802259887 using {'max_depth': 4, 'n_estimators': 100}\nTest accuracy: 0.5666666666666667\n" ], [ "randomForestModel = RandomForestClassifier(n_estimators=100,max_depth=4)\n\n#cross-val score\nscore=cross_val_score(randomForestModel, Xnew, y, cv=20)", "_____no_output_____" ], [ "randomForestModel.fit(x_train_new,np.ravel(y_train_new))\nprint('Training score: ', randomForestModel.score(x_train_new,np.ravel(y_train_new)))\nprint('Test score: ', randomForestModel.score(x_test_new,np.ravel(y_test_new)))", "Training score: 0.7364016736401674\nTest score: 0.6166666666666667\n" ] ] ]

[ "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import pandas as pd\ndf = pd.read_csv(\"../data/df_baseline.csv\")", "_____no_output_____" ] ], [ [ "## Load Data", "_____no_output_____" ] ], [ [ "print(df.shape)\ndf.head()", "(1083, 18)\n" ], [ "df.info()", "<class 'pandas.core.frame.DataFrame'>\nRangeIndex: 1083 entries, 0 to 1082\nData columns (total 18 columns):\nlabel 1083 non-null int64\nartist 1083 non-null object\nalbum 1083 non-null object\ngenre 1083 non-null object\nsingle_count 1083 non-null int64\nfreq_billboard 1083 non-null int64\nfreq_genius 1083 non-null int64\nfreq_theSource 1083 non-null int64\nfreq_xxl 1083 non-null int64\nrating_AOTY 61 non-null float64\nrating_meta 324 non-null float64\nrating_pitch 220 non-null float64\ntwitter 1083 non-null int64\ninstagram 1083 non-null int64\nfacebook 1083 non-null int64\nspotify 1083 non-null int64\nsoundcloud 1083 non-null int64\nyoutube 1083 non-null int64\ndtypes: float64(3), int64(12), object(3)\nmemory usage: 152.4+ KB\n" ] ], [ [ "**Note**\n- 온라인매체 기사의 양, 평론가 평점은 Null Value가 있기 때문에, 당장 Decision Tree를 통해 학습을 시킬 수 없어, Feature에서 제외를 한다.", "_____no_output_____" ], [ "## Data Preparation for Modeling", "_____no_output_____" ], [ "#### 장르 `hiphop`, `R&B`, `Soul`, `Funk`, `Pop`", "_____no_output_____" ] ], [ [ "df = pd.get_dummies(df, columns=['genre'])", "_____no_output_____" ], [ "df.columns", "_____no_output_____" ] ], [ [ "#### Split train & test data ", "_____no_output_____" ] ], [ [ "feature_names = ['single_count', 'freq_billboard',\n 'freq_genius', 'freq_theSource', 'freq_xxl',\n 'twitter', 'instagram', 'facebook',\n 'spotify', 'soundcloud', 'youtube',\n 'genre_funk', 'genre_hiphop', 'genre_pop', 'genre_rnb', 'genre_soul']\ndfX = df[feature_names].copy()\ndfy = df['label'].copy()", "_____no_output_____" ], [ "dfX.tail()", "_____no_output_____" ], [ "dfy.tail()", "_____no_output_____" ], [ "from sklearn.model_selection import train_test_split\nX_train, X_test, y_train, y_test = train_test_split(dfX, dfy, test_size=0.25, random_state=0)", "_____no_output_____" ] ], [ [ "# KNN", "_____no_output_____" ] ], [ [ "from sklearn.neighbors import KNeighborsClassifier", "_____no_output_____" ], [ "model = KNeighborsClassifier(n_neighbors=10).fit(X_train, y_train)", "_____no_output_____" ], [ "from sklearn.metrics import confusion_matrix", "_____no_output_____" ], [ "confusion_matrix(y_test, model.predict(X_test))", "_____no_output_____" ], [ "from sklearn.metrics import classification_report", "_____no_output_____" ], [ "print(classification_report(y_test, model.predict(X_test)))", " precision recall f1-score support\n\n 0 0.85 0.97 0.91 210\n 1 0.79 0.43 0.55 61\n\navg / total 0.84 0.85 0.83 271\n\n" ], [ "from sklearn.metrics import roc_curve\nimport matplotlib.pyplot as plt\n\nfpr, tpr, thresholds = roc_curve(y_test, model.predict_proba(X_test)[:, 1])\n\nplt.plot(fpr, tpr, label=\"KNN\")\nplt.legend()\nplt.plot([0, 1], [0, 1], 'k--', label=\"random guess\")\nplt.xlabel('False Positive Rate (Fall-Out)')\nplt.ylabel('True Positive Rate (Recall)')\nplt.title('Receiver operating characteristic example')\nplt.show()", "_____no_output_____" ], [ "from sklearn.metrics import auc\nauc(fpr, tpr)", "_____no_output_____" ] ], [ [ "# SGD", "_____no_output_____" ] ], [ [ "from sklearn.linear_model import SGDClassifier", "_____no_output_____" ], [ "model_SGD = SGDClassifier(random_state=0).fit(X_train, y_train)", "/Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/sklearn/linear_model/stochastic_gradient.py:128: FutureWarning: max_iter and tol parameters have been added in <class 'sklearn.linear_model.stochastic_gradient.SGDClassifier'> in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n \"and default tol will be 1e-3.\" % type(self), FutureWarning)\n" ], [ "confusion_matrix(y_train, model_SGD.predict(X_train))", "_____no_output_____" ], [ "confusion_matrix(y_test, model_SGD.predict(X_test))", "_____no_output_____" ], [ "print(classification_report(y_test, model_SGD.predict(X_test)))", " precision recall f1-score support\n\n 0 0.94 0.40 0.57 210\n 1 0.31 0.92 0.46 61\n\navg / total 0.80 0.52 0.54 271\n\n" ], [ "fpr, tpr, thresholds = roc_curve(y_test, model_SGD.predict(X_test))", "_____no_output_____" ], [ "plt.figure(figsize=(10, 10))\n\nplt.plot(fpr, tpr, label=\"roc curve\")\nplt.legend()\nplt.plot([0, 1], [0, 1], 'k--', label=\"random guess\")\nplt.xlabel('False Positive Rate (Fall-Out)')\nplt.ylabel('True Positive Rate (Recall)')\nplt.title('Receiver operating characteristic example')\nplt.show()", "_____no_output_____" ], [ "auc(fpr, tpr)", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "empty" ] ] ]

[ "empty" ]

[ [ "empty" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<img src=\"../../img/logo_amds.png\" alt=\"Logo\" style=\"width: 128px;\"/>\n\n# AmsterdamUMCdb - Freely Accessible ICU Database\n\nversion 1.0.2 March 2020 \nCopyright &copy; 2003-2020 Amsterdam UMC - Amsterdam Medical Data Science", "_____no_output_____" ], [ "# Vasopressors and inotropes\nShows medication for artificially increasing blood pressure (vasopressors) or stimulating heart function (inotropes), if any, a patient received.", "_____no_output_____" ], [ "## Imports", "_____no_output_____" ] ], [ [ "%matplotlib inline\nimport amsterdamumcdb\nimport psycopg2\nimport pandas as pd\nimport numpy as np\nimport re\n\nimport matplotlib.pyplot as plt\nimport matplotlib.gridspec as gridspec\nimport matplotlib as mpl\n\nimport io\nfrom IPython.display import display, HTML, Markdown", "_____no_output_____" ] ], [ [ "## Display settings", "_____no_output_____" ] ], [ [ "#matplotlib settings for image size\n#needs to be in a different cell from %matplotlib inline\nplt.style.use('seaborn-darkgrid')\nplt.rcParams[\"figure.dpi\"] = 288\nplt.rcParams[\"figure.figsize\"] = [16, 12]\nplt.rcParams[\"font.size\"] = 12\n\npd.options.display.max_columns = None\npd.options.display.max_rows = None\npd.options.display.max_colwidth = 1000", "_____no_output_____" ] ], [ [ "## Connection settings", "_____no_output_____" ] ], [ [ "#Modify config.ini in the root folder of the repository to change the settings to connect to your postgreSQL database\nimport configparser\nimport os\nconfig = configparser.ConfigParser()\n\nif os.path.isfile('../../config.ini'):\n config.read('../../config.ini')\nelse:\n config.read('../../config.SAMPLE.ini')\n\n#Open a connection to the postgres database:\ncon = psycopg2.connect(database=config['psycopg2']['database'], \n user=config['psycopg2']['username'], password=config['psycopg2']['password'], \n host=config['psycopg2']['host'], port=config['psycopg2']['port'])\ncon.set_client_encoding('WIN1252') #Uses code page for Dutch accented characters.\ncon.set_session(autocommit=True)\n\ncursor = con.cursor()\ncursor.execute('SET SCHEMA \\'amsterdamumcdb\\''); #set search_path to amsterdamumcdb schema", "_____no_output_____" ] ], [ [ "## Vasopressors and inotropes\nfrom drugitems", "_____no_output_____" ] ], [ [ "sql_vaso_ino = \"\"\"\nWITH vasopressor_inotropes AS (\n SELECT\n admissionid,\n CASE \n WHEN COUNT(*) > 0 THEN TRUE\n ELSE FALSE\n END AS vasopressors_inotropes_bool,\n STRING_AGG(DISTINCT item, '; ') AS vasopressors_inotropes_given\n FROM drugitems\n WHERE \n ordercategoryid = 65 -- continuous i.v. perfusor\n AND itemid IN (\n 6818, -- Adrenaline (Epinefrine)\n 7135, -- Isoprenaline (Isuprel)\n 7178, -- Dobutamine (Dobutrex)\n 7179, -- Dopamine (Inotropin)\n 7196, -- Enoximon (Perfan)\n 7229, -- Noradrenaline (Norepinefrine)\n 12467, -- Terlipressine (Glypressin)\n 13490, -- Methyleenblauw IV (Methylthionide cloride)\n 19929 -- Fenylefrine\n )\n AND rate > 0.1\n GROUP BY admissionid\n)\nSELECT\n a.admissionid, location, \n CASE \n WHEN vi.vasopressors_inotropes_bool Then TRUE\n ELSE FALSE\n END AS vasopressors_inotropes_bool,\n vasopressors_inotropes_given\nFROM admissions a\nLEFT JOIN vasopressor_inotropes vi ON\n a.admissionid = vi.admissionid\n\"\"\"\nvaso_ino = pd.read_sql(sql_vaso_ino,con)\nvaso_ino.tail()", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/EvenSol/NeqSim-Colab/blob/master/notebooks/process/masstransferMeOH.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ] ], [ [ "#@title Calculation of mass transfer and hydrate inhibition of a wet gas by injection of methanol\n#@markdown Demonstration of mass transfer calculation using the NeqSim software in Python\n#@markdown <br><br>This document is part of the module [\"Introduction to Gas Processing using NeqSim in Colab\"](https://colab.research.google.com/github/EvenSol/NeqSim-Colab/blob/master/notebooks/examples_of_NeqSim_in_Colab.ipynb#scrollTo=_eRtkQnHpL70).\n%%capture\n!pip install neqsim\nimport neqsim\nfrom neqsim.thermo.thermoTools import *\nimport matplotlib\nimport numpy as np\nimport matplotlib.pyplot as plt\nfrom neqsim.thermo import fluid, fluid_df\nimport pandas as pd\nfrom neqsim.process import gasscrubber, clearProcess, run,nequnit, phasemixer, splitter, clearProcess, stream, valve, separator, compressor, runProcess, viewProcess, heater,saturator, mixer\n\nplt.style.use('classic')\n%matplotlib inline", "_____no_output_____" ] ], [ [ "#Mass transfer calculations\nModel for mass transfer calculation in NeqSim based on Solbraa (2002):\n\nhttps://ntnuopen.ntnu.no/ntnu-xmlui/handle/11250/231326", "_____no_output_____" ], [ "In the following calculations we assume a water saturated gas the is mixed with pure liquid methanol. These phases are not in equiibrium when they enter the pipeline. When the gas and methanol liquid comes in contact in the pipeline, methanol will vaporize into the gas, and water (and other comonents from the gas) will be absorbed into the liquid methanol. The focus of the following calculations will be to evaluate the mass transfer as function of contanct length with gas and methanol. It also evaluates the hydrate temperature of the gas leaving the pipe section.\n\n\n\n\n", "_____no_output_____" ], [ "Figure 1 Illustration of mass transfer process\n\n", "_____no_output_____" ], [ "**The parameters for the model are:**\n\nTemperature and pressure of the pipe (mass transfer calculated at constant temperature and pressure).\n\nLength and diameter of pipe where gas and liquid will be in contact and mass transfer can occur.\n\nFlow rate of the gas in MSm3/day, flow rate of methanol (kg/hr).", "_____no_output_____" ], [ "#Calculation of compostion of aqueous phase and gas leaving pipe section\nIn the following script we will simulate the composition of the gas leaving pipe section at a given pipe lengt.", "_____no_output_____" ] ], [ [ "# Input parameters\npressure = 52.21 # bara\ntemperature = 15.2 #C\ngasFlow = 1.23 #MSm3/day\nmethanolFlow = 6000.23 # kg/day\n\npipelength = 10.0 #meter\npipeInnerDiameter = 0.5 #meter\n\n# Create a gas-condensate fluid\nfeedgas = {'ComponentName': [\"nitrogen\",\"CO2\",\"methane\", \"ethane\" , \"propane\", \"i-butane\", \"n-butane\", \"water\", \"methanol\"], \n 'MolarComposition[-]': [0.01, 0.01, 0.8, 0.06, 0.01,0.005,0.005, 0.0, 0.0]\n} \n\nnaturalgasFluid = fluid_df(pd.DataFrame(feedgas)).setModel(\"CPAs-SRK-EOS-statoil\")\nnaturalgasFluid.setTotalFlowRate(gasFlow, \"MSm3/day\")\nnaturalgasFluid.setTemperature(temperature, \"C\")\nnaturalgasFluid.setPressure(pressure, \"bara\")\n\n# Create a liquid methanol fluid\nfeedMeOH = {'ComponentName': [\"nitrogen\",\"CO2\",\"methane\", \"ethane\" , \"propane\", \"i-butane\", \"n-butane\", \"water\", \"methanol\"], \n 'MolarComposition[-]': [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,1.0]\n} \nmeOHFluid = fluid_df(pd.DataFrame(feedMeOH) ).setModel(\"CPAs-SRK-EOS-statoil\")\nmeOHFluid.setTotalFlowRate(methanolFlow, \"kg/hr\");\nmeOHFluid.setTemperature(temperature, \"C\");\nmeOHFluid.setPressure(pressure, \"bara\");\n\nclearProcess()\ndryinjectiongas = stream(naturalgasFluid)\nMeOHFeed = stream(meOHFluid)\nwatersaturator = saturator(dryinjectiongas)\nwaterSaturatedFeedGas = stream(watersaturator.getOutStream())\n\nmainMixer = phasemixer(\"gas MeOH mixer\")\nmainMixer.addStream(waterSaturatedFeedGas)\nmainMixer.addStream(MeOHFeed)\n\t\t\npipeline = nequnit(mainMixer.getOutStream(), equipment=\"pipeline\", flowpattern=\"stratified\") #alternative flow patterns are: stratified, annular and droplet\npipeline.setLength(pipelength)\npipeline.setID(pipeInnerDiameter)\n\nscrubber = gasscrubber(pipeline.getOutStream())\ngasFromScrubber = stream(scrubber.getGasOutStream())\naqueousFromScrubber = stream(scrubber.getLiquidOutStream())\nrun()\nprint('Composition of gas leaving pipe section after ', pipelength, ' meter')\nprintFrame(gasFromScrubber.getFluid())\nprint('Composition of aqueous phase leaving pipe section after ', pipelength, ' meter')\nprintFrame(aqueousFromScrubber.getFluid())\n\nprint('Interface contact area ', pipeline.getInterfacialArea(), ' m^2')\nprint('Volume fraction aqueous phase ', pipeline.getOutStream().getFluid().getVolumeFraction(1), ' -')", "Composition of gas leaving pipe section after 10.0 meter\n total gas \n nitrogen 1.11418E-2 1.11418E-2 [mole fraction]\n CO2 1.08558E-2 1.08558E-2 [mole fraction]\n methane 8.89321E-1 8.89321E-1 [mole fraction]\n ethane 6.59386E-2 6.59386E-2 [mole fraction]\n propane 1.10391E-2 1.10391E-2 [mole fraction]\n i-butane 5.4792E-3 5.4792E-3 [mole fraction]\n n-butane 5.48529E-3 5.48529E-3 [mole fraction]\n water 3.42883E-4 3.42883E-4 [mole fraction]\n methanol 3.96384E-4 3.96384E-4 [mole fraction]\n \n Density 4.51425E1 [kg/m^3]\n PhaseFraction 1E0 [mole fraction]\n MolarMass 1.8183E1 1.8183E1 [kg/kmol]\n Z factor 8.7716E-1 [-]\n Heat Capacity (Cp) 2.54484E0 [kJ/kg*K]\n Heat Capacity (Cv) 1.65968E0 [kJ/kg*K]\n Speed of Sound 3.96268E2 [m/sec]\n Enthalpy -3.29472E1 -3.29472E1 [kJ/kg]\n Entropy -1.62974E0 -1.62974E0 [kJ/kg*K]\n JT coefficient 5.03748E-1 [K/bar]\n \n Viscosity 1.22106E-5 [kg/m*sec]\n Conductivity 3.74742E-2 [W/m*K]\n SurfaceTension [N/m]\n \n \n \n Pressure 52.21 [bar]\n Temperature 288.34999999999997 [K]\n \n Model CPAs-SRK-EOS-statoil -\n Mixing Rule classic-CPA_T -\n \n Stream -\n \n \n \n \nComposition of aqueous phase leaving pipe section after 10.0 meter\n total aqueous \n nitrogen 1.53381E-4 1.53381E-4 [mole fraction]\n CO2 3.28961E-3 3.28961E-3 [mole fraction]\n methane 3.44647E-2 3.44647E-2 [mole fraction]\n ethane 1.09255E-2 1.09255E-2 [mole fraction]\n propane 1.28029E-3 1.28029E-3 [mole fraction]\n i-butane 1.08241E-3 1.08241E-3 [mole fraction]\n n-butane 1.01559E-3 1.01559E-3 [mole fraction]\n water 8.18682E-4 8.18682E-4 [mole fraction]\n methanol 9.4697E-1 9.4697E-1 [mole fraction]\n \n Density 7.82709E2 [kg/m^3]\n PhaseFraction 1E0 [mole fraction]\n MolarMass 3.15665E1 3.15665E1 [kg/kmol]\n Z factor 8.7826E-2 [-]\n Heat Capacity (Cp) 2.26412E0 [kJ/kg*K]\n Heat Capacity (Cv) 1.885E0 [kJ/kg*K]\n Speed of Sound 1.07486E3 [m/sec]\n Enthalpy -1.14087E3 -1.14087E3 [kJ/kg]\n Entropy -3.37618E0 -3.37618E0 [kJ/kg*K]\n JT coefficient -3.74052E-2 [K/bar]\n \n Viscosity 5.85317E-4 [kg/m*sec]\n Conductivity 5.89686E-1 [W/m*K]\n SurfaceTension [N/m]\n \n \n \n Pressure 52.21 [bar]\n Temperature 288.34999999999997 [K]\n \n Model CPAs-SRK-EOS-statoil -\n Mixing Rule classic-CPA_T -\n \n Stream -\n \n \n \n \nInterface contact area 2.641129854675618 m^2\nVolume fraction aqueous phase 0.011201474850165916 -\n" ] ], [ [ "# Calculation of hydrate equilibrium temperature of gas leaving pipe section\nIn the following script we will simulate the composition of the gas leaving pipe section as well as hydrate equilibrium temperature of this gas as function of pipe length.", "_____no_output_____" ] ], [ [ "maxpipelength = 10.0\n\ndef hydtemps(length):\n pipeline.setLength(length)\n run();\n return gasFromScrubber.getHydrateEquilibriumTemperature()-273.15\n\nlength = np.arange(0.01, maxpipelength, (maxpipelength)/10.0)\nhydtem = [hydtemps(length2) for length2 in length]\n\nplt.figure()\nplt.plot(length, hydtem)\nplt.xlabel('Length available for mass transfer [m]')\nplt.ylabel('Hydrate eq.temperature [C]')\nplt.title('Hydrate eq.temperature of gas leaving pipe section')", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT-0" ]

[ "MIT-0" ]

[ "MIT-0" ]

[ [ [ "# Amazon Forecast: predicting time-series at scale\n\nForecasting is used in a variety of applications and business use cases: For example, retailers need to forecast the sales of their products to decide how much stock they need by location, Manufacturers need to estimate the number of parts required at their factories to optimize their supply chain, Businesses need to estimate their flexible workforce needs, Utilities need to forecast electricity consumption needs in order to attain an efficient energy network, and \nenterprises need to estimate their cloud infrastructure needs.", "_____no_output_____" ], [ "<img src=\"BlogImages/amazon_forecast.png\">\n\n\n# Table of Contents\n\n* Step 0: [Setting up](#setup)\n* Step 1: [Preparing the Datasets](#prepare)\n* Step 2: [Importing the Data](#import)\n * Step 2a: [Creating a Dataset Group](#create)\n * Step 2b: [Creating a Target Dataset](#target)\n * Step 2c: [Creating a Related Dataset](#related)\n * Step 2d: [Update the Dataset Group](#update)\n * Step 2e: [Creating a Target Time Series Dataset Import Job](#targetImport)\n * Step 2f: [Creating a Related Time Series Dataset Import Job](#relatedImport)\n* Step 3: [Choosing an Algorithm and Evaluating its Performance](#algo)\n * Step 3a: [Choosing DeepAR+](#DeepAR)\n * Step 3b: [Choosing Prophet](#prophet)\n* Step 4: [Computing Error Metrics from Backtesting](#error)\n* Step 5: [Creating a Forecast](#forecast)\n* Step 6: [Querying the Forecasts](#query)\n* Step 7: [Exporting the Forecasts](#export)\n* Step 8: [Clearning up your Resources](#cleanup)\n", "_____no_output_____" ], [ "# First let us setup Amazon Forecast<a class=\"anchor\" id=\"setup\">\n\nThis section sets up the permissions and relevant endpoints.", "_____no_output_____" ] ], [ [ "%load_ext autoreload\n%autoreload 2\nfrom util.fcst_utils import *\nimport warnings\nimport boto3\nimport s3fs\nplt.rcParams['figure.figsize'] = (15.0, 5.0)\nwarnings.filterwarnings('ignore')", "_____no_output_____" ] ], [ [ "Although, we have set the region to us-west-2 below, you can choose any of the 6 regions that the service is available in.", "_____no_output_____" ] ], [ [ "region = 'us-west-2'\nbucket = 'bike-demo'\nversion = 'prod'", "_____no_output_____" ], [ "session = boto3.Session(region_name=region) \nforecast = session.client(service_name='forecast') \nforecast_query = session.client(service_name='forecastquery')", "_____no_output_____" ], [ "role_arn = get_or_create_role_arn()", "_____no_output_____" ] ], [ [ "# Overview\n\n<img src=\"BlogImages/outline.png\">\n\n<img src=\"BlogImages/forecast_workflow.png\">\n\nThe above figure summarizes the key workflow of using Forecast. ", "_____no_output_____" ], [ "# Step 1: Preparing the Datasets<a class=\"anchor\" id=\"prepare\">", "_____no_output_____" ] ], [ [ "bike_df = pd.read_csv(\"../data/train.csv\", dtype = object)\nbike_df.head()", "_____no_output_____" ], [ "bike_df['count'] = bike_df['count'].astype('float')\nbike_df['workingday'] = bike_df['workingday'].astype('float')", "_____no_output_____" ] ], [ [ "We take about two and a half week's of hourly data for demonstration, just for the purpose that there's no missing data in the whole range.", "_____no_output_____" ] ], [ [ "bike_df_small = bike_df[-2*7*24-24*3:]\nbike_df_small['item_id'] = \"bike_12\"", "_____no_output_____" ] ], [ [ "Let us plot the time series first.", "_____no_output_____" ] ], [ [ "bike_df_small.plot(x='datetime', y='count', figsize=(15, 8))", "_____no_output_____" ] ], [ [ "We can see that the target time series seem to have a drop over weekends. Next let's plot both the target time series and the related time series that indicates whether today is a `workday` or not. More precisely, $r_t = 1$ if $t$ is a work day and 0 if not.", "_____no_output_____" ] ], [ [ "plt.figure(figsize=(15, 8))\nax = plt.gca()\nbike_df_small.plot(x='datetime', y='count', ax=ax);\nax2 = ax.twinx()\nbike_df_small.plot(x='datetime', y='workingday', color='red', ax=ax2);", "_____no_output_____" ] ], [ [ "Notice that to use the related time series, we need to ensure that the related time series covers the whole target time series, as well as the future values as specified by the forecast horizon. More precisely, we need to make sure:\n```\nlen(related time series) >= len(target time series) + forecast horizon\n```\nBasically, all items need to have data start at or before the item start date, and have data until the forecast horizon (i.e. the latest end date across all items + forecast horizon). Additionally, there should be no missing values in the related time series. The following picture illustrates the desired logic. \n\n<img src=\"BlogImages/rts_viz.png\">\n\nFor more details regarding how to prepare your Related Time Series dataset, please refer to the public documentation <a href=\"https://docs.aws.amazon.com/forecast/latest/dg/related-time-series-datasets.html\">here</a>. \n\nSuppose in this particular example, we wish to forecast for the next 24 hours, and thus we generate the following dataset.", "_____no_output_____" ] ], [ [ "target_df = bike_df_small[['item_id', 'datetime', 'count']][:-24]\nrts_df = bike_df_small[['item_id', 'datetime', 'workingday']]", "_____no_output_____" ], [ "target_df.head(5)", "_____no_output_____" ] ], [ [ "As we can see, the length of the related time series is equal to the length of the target time series plus the forecast horizon. ", "_____no_output_____" ] ], [ [ "print(len(target_df), len(rts_df))\nassert len(target_df) + 24 == len(rts_df), \"length doesn't match\"", "_____no_output_____" ] ], [ [ "Next we check whether there are \"holes\" in the related time series. ", "_____no_output_____" ] ], [ [ "assert len(rts_df) == len(pd.date_range(\n start=list(rts_df['datetime'])[0],\n end=list(rts_df['datetime'])[-1],\n freq='H'\n)), \"missing entries in the related time series\"", "_____no_output_____" ] ], [ [ "Everything looks fine, and we plot both time series again. As it can be seen, the related time series (indicator of whether the current day is a workday or not) is longer than the target time series. The binary working day indicator feature is a good example of a related time series, since it is known at all future time points. Other examples of related time series include holiday and promotion features.", "_____no_output_____" ] ], [ [ "plt.figure(figsize=(15, 10))\nax = plt.gca()\ntarget_df.plot(x='datetime', y='count', ax=ax);\nax2 = ax.twinx()\nrts_df.plot(x='datetime', y='workingday', color='red', ax=ax2);", "_____no_output_____" ], [ "target_df.to_csv(\"../data/bike_small.csv\", index= False, header = False)\nrts_df.to_csv(\"../data/bike_small_rts.csv\", index= False, header = False)", "_____no_output_____" ], [ "s3 = session.client('s3')\naccount_id = boto3.client('sts').get_caller_identity().get('Account')", "_____no_output_____" ] ], [ [ "If you don't have this bucket `amazon-forecast-data-{account_id}`, create it first on S3.", "_____no_output_____" ] ], [ [ "bucket_name = f\"amazon-forecast-data-{account_id}\"\nkey = \"bike_small\"\n\ns3.upload_file(Filename=\"../data/bike_small.csv\", Bucket = bucket_name, Key = f\"{key}/bike.csv\")\ns3.upload_file(Filename=\"../data/bike_small_rts.csv\", Bucket = bucket_name, Key = f\"{key}/bike_rts.csv\")", "_____no_output_____" ] ], [ [ "# Step 2. Importing the Data<a class=\"anchor\" id=\"import\">\n\n\nNow we are ready to import the datasets into the Forecast service. Starting from the raw data, Amazon Forecast automatically extracts the dataset that is suitable for forecasting. As an example, a retailer normally records the transaction record such as\n<img src=\"BlogImages/data_format.png\">\n<img src=\"BlogImages/timestamp.png\">\n\n", "_____no_output_____" ] ], [ [ "project = \"bike_rts_demo\"\nidx = 4", "_____no_output_____" ], [ "s3_data_path = f\"s3://{bucket_name}/{key}\"", "_____no_output_____" ] ], [ [ "Below, we specify key input data and forecast parameters", "_____no_output_____" ] ], [ [ "freq = \"H\"\nforecast_horizon = 24\ntimestamp_format = \"yyyy-MM-dd HH:mm:ss\"\ndelimiter = ','", "_____no_output_____" ] ], [ [ "## Step 2a. Creating a Dataset Group<a class=\"anchor\" id=\"create\">\nFirst let's create a dataset group and then update it later to add our datasets.", "_____no_output_____" ] ], [ [ "dataset_group = f\"{project}_gp_{idx}\"\ndataset_arns = []\ncreate_dataset_group_response = forecast.create_dataset_group(Domain=\"RETAIL\",\n DatasetGroupName=dataset_group,\n DatasetArns=dataset_arns)", "_____no_output_____" ], [ "logging.info(f'Creating dataset group {dataset_group}')", "_____no_output_____" ], [ "dataset_group_arn = create_dataset_group_response['DatasetGroupArn']", "_____no_output_____" ], [ "forecast.describe_dataset_group(DatasetGroupArn=dataset_group_arn)", "_____no_output_____" ] ], [ [ "## Step 2b. Creating a Target Dataset<a class=\"anchor\" id=\"target\">\nIn this example, we will define a target time series. This is a required dataset to use the service.", "_____no_output_____" ], [ "Below we specify the target time series name af_demo_ts_4.", "_____no_output_____" ] ], [ [ "ts_dataset_name = f\"{project}_ts_{idx}\"\nprint(ts_dataset_name)", "_____no_output_____" ] ], [ [ "Next, we specify the schema of our dataset below. Make sure the order of the attributes (columns) matches the raw data in the files. We follow the same three attribute format as the above example.", "_____no_output_____" ] ], [ [ "ts_schema_val = [{\"AttributeName\": \"item_id\", \"AttributeType\": \"string\"},\n {\"AttributeName\": \"timestamp\", \"AttributeType\": \"timestamp\"},\n {\"AttributeName\": \"demand\", \"AttributeType\": \"float\"}]\nts_schema = {\"Attributes\": ts_schema_val}", "_____no_output_____" ], [ "logging.info(f'Creating target dataset {ts_dataset_name}')", "_____no_output_____" ], [ "response = forecast.create_dataset(Domain=\"RETAIL\",\n DatasetType='TARGET_TIME_SERIES',\n DatasetName=ts_dataset_name,\n DataFrequency=freq,\n Schema=ts_schema\n )", "_____no_output_____" ], [ "ts_dataset_arn = response['DatasetArn']", "_____no_output_____" ], [ "forecast.describe_dataset(DatasetArn=ts_dataset_arn)", "_____no_output_____" ] ], [ [ "## Step 2c. Creating a Related Dataset<a class=\"anchor\" id=\"related\">\nIn this example, we will define a related time series.", "_____no_output_____" ], [ "Specify the related time series name af_demo_rts_4.", "_____no_output_____" ] ], [ [ "rts_dataset_name = f\"{project}_rts_{idx}\"\nprint(rts_dataset_name)", "_____no_output_____" ] ], [ [ "Specify the schema of your dataset here. Make sure the order of columns matches the raw data files. We follow the same three column format as the above example.", "_____no_output_____" ] ], [ [ "rts_schema_val = [{\"AttributeName\": \"item_id\", \"AttributeType\": \"string\"},\n {\"AttributeName\": \"timestamp\", \"AttributeType\": \"timestamp\"},\n {\"AttributeName\": \"price\", \"AttributeType\": \"float\"}]\nrts_schema = {\"Attributes\": rts_schema_val}", "_____no_output_____" ], [ "logging.info(f'Creating related dataset {rts_dataset_name}')", "_____no_output_____" ], [ "response = forecast.create_dataset(Domain=\"RETAIL\",\n DatasetType='RELATED_TIME_SERIES',\n DatasetName=rts_dataset_name,\n DataFrequency=freq,\n Schema=rts_schema\n )", "_____no_output_____" ], [ "rts_dataset_arn = response['DatasetArn']", "_____no_output_____" ], [ "forecast.describe_dataset(DatasetArn=rts_dataset_arn)", "_____no_output_____" ] ], [ [ "## Step 2d. Updating the dataset group with the datasets we created<a class=\"anchor\" id=\"update\">\nYou can have multiple datasets under the same dataset group. Update it with the datasets we created before.", "_____no_output_____" ] ], [ [ "dataset_arns = []\ndataset_arns.append(ts_dataset_arn)\ndataset_arns.append(rts_dataset_arn)\nforecast.update_dataset_group(DatasetGroupArn=dataset_group_arn, DatasetArns=dataset_arns)", "_____no_output_____" ], [ "forecast.describe_dataset_group(DatasetGroupArn=dataset_group_arn)", "_____no_output_____" ] ], [ [ "## Step 2e. Creating a Target Time Series Dataset Import Job<a class=\"anchor\" id=\"targetImport\">", "_____no_output_____" ] ], [ [ "ts_s3_data_path = f\"{s3_data_path}/bike.csv\"", "_____no_output_____" ], [ "ts_dataset_import_job_response = forecast.create_dataset_import_job(DatasetImportJobName=dataset_group,\n DatasetArn=ts_dataset_arn,\n DataSource= {\n \"S3Config\" : {\n \"Path\": ts_s3_data_path,\n \"RoleArn\": role_arn\n } \n },\n TimestampFormat=timestamp_format)", "_____no_output_____" ], [ "ts_dataset_import_job_arn=ts_dataset_import_job_response['DatasetImportJobArn']", "_____no_output_____" ], [ "status = wait(lambda: forecast.describe_dataset_import_job(DatasetImportJobArn=ts_dataset_import_job_arn))\nassert status", "_____no_output_____" ] ], [ [ "## Step 2f. Creating a Related Time Series Dataset Import Job<a class=\"anchor\" id=\"relatedImport\">", "_____no_output_____" ] ], [ [ "rts_s3_data_path = f\"{s3_data_path}/bike_rts.csv\"", "_____no_output_____" ], [ "rts_dataset_import_job_response = forecast.create_dataset_import_job(DatasetImportJobName=dataset_group,\n DatasetArn=rts_dataset_arn,\n DataSource= {\n \"S3Config\" : {\n \"Path\": rts_s3_data_path,\n \"RoleArn\": role_arn\n } \n },\n TimestampFormat=timestamp_format)", "_____no_output_____" ], [ "rts_dataset_import_job_arn=rts_dataset_import_job_response['DatasetImportJobArn']", "_____no_output_____" ], [ "status = wait(lambda: forecast.describe_dataset_import_job(DatasetImportJobArn=rts_dataset_import_job_arn))\nassert status", "_____no_output_____" ] ], [ [ "# Step 3. Choosing an algorithm and evaluating its performance<a class=\"anchor\" id=\"algo\">\n\nOnce the datasets are specified with the corresponding schema, Amazon Forecast will automatically aggregate all the relevant pieces of information for each item, such as sales, price, promotions, as well as categorical attributes, and generate the desired dataset. Next, one can choose an algorithm (forecasting model) and evaluate how well this particular algorithm works on this dataset. The following graph gives a high-level overview of the forecasting models.\n<img src=\"BlogImages/recipes.png\">\n<img src=\"BlogImages/pred_details.png\">\n\n\nAmazon Forecast provides several state-of-the-art forecasting algorithms including classic forecasting methods such as ETS, ARIMA, Prophet and deep learning approaches such as DeepAR+. Classical forecasting methods, such as Autoregressive Integrated Moving Average (ARIMA) or Exponential Smoothing (ETS), fit a single model to each individual time series, and then use that model to extrapolate the time series into the future. Amazon's Non-Parametric Time Series (NPTS) forecaster also fits a single model to each individual time series. Unlike the naive or seasonal naive forecasters that use a fixed time index (the previous index $T-1$ or the past season $T - \\tau$) as the prediction for time step $T$, NPTS randomly samples a time index $t \\in \\{0, \\dots T-1\\}$ in the past to generate a sample for the current time step $T$.\n\nIn many applications, you may encounter many similar time series across a set of cross-sectional units. Examples of such time series groupings are demand for different products, server loads, and requests for web pages. In this case, it can be beneficial to train a single model jointly over all of these time series. DeepAR+ takes this approach, outperforming the standard ARIMA and ETS methods when your dataset contains hundreds of related time series. The trained model can also be used for generating forecasts for new time series that are similar to the ones it has been trained on. While deep learning approaches can outperform standard methods, this is only possible when there is sufficient data available for training. It is not true for example when one trains a neural network with a time-series contains only a few dozens of observations. Amazon Forecast provides the best of two worlds allowing users to either choose a specific algorithm or let Amazon Forecast automatically perform model selection. \n\n## How to evaluate a forecasting model?\n\nBefore moving forward, let's first introduce the notion of *backtest* when evaluating forecasting models. The key difference between evaluating forecasting algorithms and standard ML applications is that we need to make sure there is no future information gets used in the past. In other words, the procedure needs to be causal. \n\n<img src=\"BlogImages/backtest.png\">\n\n\nIn this notebook, let's compare the neural network based method, DeepAR+ with Facebook's open-source Bayesian method Prophet. \n", "_____no_output_____" ] ], [ [ "algorithm_arn = 'arn:aws:forecast:::algorithm/'", "_____no_output_____" ] ], [ [ "## Step 3a. Choosing DeepAR+<a class=\"anchor\" id=\"DeepAR\">", "_____no_output_____" ] ], [ [ "algorithm = 'Deep_AR_Plus'\nalgorithm_arn_deep_ar_plus = algorithm_arn + algorithm\npredictor_name_deep_ar = f'{project}_{algorithm.lower()}_{idx}'", "_____no_output_____" ], [ "logging.info(f'[{predictor_name_deep_ar}] Creating predictor {predictor_name_deep_ar} ...')", "_____no_output_____" ], [ "create_predictor_response = forecast.create_predictor(PredictorName=predictor_name_deep_ar,\n AlgorithmArn=algorithm_arn_deep_ar_plus,\n ForecastHorizon=forecast_horizon,\n PerformAutoML=False,\n PerformHPO=False,\n InputDataConfig= {\"DatasetGroupArn\": dataset_group_arn},\n FeaturizationConfig= {\"ForecastFrequency\": freq}\n )", "_____no_output_____" ], [ "predictor_arn_deep_ar = create_predictor_response['PredictorArn']", "_____no_output_____" ], [ "status = wait(lambda: forecast.describe_predictor(PredictorArn=predictor_arn_deep_ar))\nassert status", "_____no_output_____" ], [ "forecast.describe_predictor(PredictorArn=predictor_arn_deep_ar)", "_____no_output_____" ] ], [ [ "## Step 3b. Choosing Prophet<a class=\"anchor\" id=\"prophet\">", "_____no_output_____" ] ], [ [ "algorithm = 'Prophet'\nalgorithm_arn_prophet = algorithm_arn + algorithm\npredictor_name_prophet = f'{project}_{algorithm.lower()}_{idx}'", "_____no_output_____" ], [ "algorithm_arn_prophet", "_____no_output_____" ], [ "logging.info(f'[{predictor_name_prophet}] Creating predictor %s ...' % predictor_name_prophet)", "_____no_output_____" ], [ "create_predictor_response = forecast.create_predictor(PredictorName=predictor_name_prophet,\n AlgorithmArn=algorithm_arn_prophet,\n ForecastHorizon=forecast_horizon,\n PerformAutoML=False,\n PerformHPO=False,\n InputDataConfig= {\"DatasetGroupArn\": dataset_group_arn},\n FeaturizationConfig= {\"ForecastFrequency\": freq}\n )", "_____no_output_____" ], [ "predictor_arn_prophet = create_predictor_response['PredictorArn']", "_____no_output_____" ], [ "status = wait(lambda: forecast.describe_predictor(PredictorArn=predictor_arn_prophet))\nassert status", "_____no_output_____" ], [ "forecast.describe_predictor(PredictorArn=predictor_arn_prophet)", "_____no_output_____" ] ], [ [ "# Step 4. Computing Error Metrics from Backtesting<a class=\"anchor\" id=\"error\">", "_____no_output_____" ], [ "After creating the predictors, we can query the forecast accuracy given by the backtest scenario and have a quantitative understanding of the performance of the algorithm. Such a process is iterative in nature during model development. When an algorithm with satisfying performance is found, the customer can deploy the predictor into a production environment, and query the forecasts for a particular item to make business decisions. The figure below shows a sample plot of different quantile forecasts of a predictor.", "_____no_output_____" ] ], [ [ "logging.info('Done creating predictor. Getting accuracy numbers for DeepAR+ ...')", "_____no_output_____" ], [ "error_metrics_deep_ar_plus = forecast.get_accuracy_metrics(PredictorArn=predictor_arn_deep_ar)\nerror_metrics_deep_ar_plus", "_____no_output_____" ], [ "logging.info('Done creating predictor. Getting accuracy numbers for Prophet ...')", "_____no_output_____" ], [ "error_metrics_prophet = forecast.get_accuracy_metrics(PredictorArn=predictor_arn_prophet)\nerror_metrics_prophet", "_____no_output_____" ], [ "def extract_summary_metrics(metric_response, predictor_name):\n df = pd.DataFrame(metric_response['PredictorEvaluationResults']\n [0]['TestWindows'][0]['Metrics']['WeightedQuantileLosses'])\n df['Predictor'] = predictor_name\n return df", "_____no_output_____" ], [ "deep_ar_metrics = extract_summary_metrics(error_metrics_deep_ar_plus, \"DeepAR\")\nprophet_metrics = extract_summary_metrics(error_metrics_prophet, \"Prophet\")", "_____no_output_____" ], [ "pd.concat([deep_ar_metrics, prophet_metrics]) \\\n .pivot(index='Quantile', columns='Predictor', values='LossValue').plot.bar();", "_____no_output_____" ] ], [ [ "As we mentioned before, if you only have a handful of time series (in this case, only 1) with a small number of examples, the neural network models (DeepAR+) are not the best choice. Here, we clearly see that DeepAR+ behaves worse than Prophet in the case of a single time series. ", "_____no_output_____" ], [ "# Step 5. Creating a Forecast<a class=\"anchor\" id=\"forecast\">\n\nNext we re-train with the full dataset, and create the forecast.", "_____no_output_____" ] ], [ [ "logging.info(f\"Done fetching accuracy numbers. Creating forecaster for DeepAR+ ...\")", "_____no_output_____" ], [ "forecast_name_deep_ar = f'{project}_deep_ar_plus_{idx}'", "_____no_output_____" ], [ "create_forecast_response_deep_ar = forecast.create_forecast(ForecastName=forecast_name_deep_ar,\n PredictorArn=predictor_arn_deep_ar)", "_____no_output_____" ], [ "forecast_arn_deep_ar = create_forecast_response_deep_ar['ForecastArn']", "_____no_output_____" ], [ "status = wait(lambda: forecast.describe_forecast(ForecastArn=forecast_arn_deep_ar))\nassert status", "_____no_output_____" ], [ "forecast.describe_forecast(ForecastArn=forecast_arn_deep_ar)", "_____no_output_____" ], [ "logging.info(f\"Done fetching accuracy numbers. Creating forecaster for Prophet ...\")", "_____no_output_____" ], [ "forecast_name_prophet = f'{project}_prophet_{idx}'", "_____no_output_____" ], [ "create_forecast_response_prophet = forecast.create_forecast(ForecastName=forecast_name_prophet,\n PredictorArn=predictor_arn_prophet)", "_____no_output_____" ], [ "forecast_arn_prophet = create_forecast_response_prophet['ForecastArn']", "_____no_output_____" ], [ "status = wait(lambda: forecast.describe_forecast(ForecastArn=forecast_arn_prophet))\nassert status", "_____no_output_____" ], [ "forecast.describe_forecast(ForecastArn=forecast_arn_prophet)", "_____no_output_____" ] ], [ [ "# Step 6. Querying the Forecasts<a class=\"anchor\" id=\"query\">", "_____no_output_____" ] ], [ [ "item_id = 'bike_12'", "_____no_output_____" ], [ "forecast_response_deep = forecast_query.query_forecast(\n ForecastArn=forecast_arn_deep_ar,\n Filters={\"item_id\": item_id})", "_____no_output_____" ], [ "forecast_response_prophet = forecast_query.query_forecast(ForecastArn=forecast_arn_prophet,\n Filters={\"item_id\":item_id})", "_____no_output_____" ], [ "fname = f'../data/bike_small.csv'\nexact = load_exact_sol(fname, item_id)", "_____no_output_____" ], [ "plot_forecasts(forecast_response_deep, exact)\nplt.title(\"DeepAR Forecast\");", "_____no_output_____" ], [ "plot_forecasts(forecast_response_prophet,exact)\nplt.title(\"Prophet Forecast\");", "_____no_output_____" ] ], [ [ "# Step 7. Exporting your Forecasts<a class=\"anchor\" id=\"export\">", "_____no_output_____" ] ], [ [ "forecast_export_name_deep_ar = f'{project}_forecast_export_deep_ar_plus_{idx}'\nforecast_export_name_deep_ar_path = f\"{s3_data_path}/{forecast_export_name_deep_ar}\"", "_____no_output_____" ], [ "create_forecast_export_response_deep_ar = forecast.create_forecast_export_job(ForecastExportJobName=forecast_export_name_deep_ar,\n ForecastArn=forecast_arn_deep_ar,\n Destination={\n \"S3Config\" : {\n \"Path\": forecast_export_name_deep_ar_path,\n \"RoleArn\": role_arn\n }\n })\nforecast_export_arn_deep_ar = create_forecast_export_response_deep_ar['ForecastExportJobArn']", "_____no_output_____" ], [ "status = wait(lambda: forecast.describe_forecast_export_job(ForecastExportJobArn = forecast_export_arn_deep_ar))\nassert status", "_____no_output_____" ], [ "forecast_export_name_prophet = f'{project}_forecast_export_prophet_{idx}'\nforecast_export_name_prophet_path = f\"{s3_data_path}/{forecast_export_name_prophet}\"", "_____no_output_____" ], [ "create_forecast_export_response_prophet = forecast.create_forecast_export_job(ForecastExportJobName=forecast_export_name_prophet,\n ForecastArn=forecast_arn_prophet,\n Destination={\n \"S3Config\" : {\n \"Path\": forecast_export_name_prophet_path,\n \"RoleArn\": role_arn\n }\n })\nforecast_export_arn_prophet = create_forecast_export_response_prophet['ForecastExportJobArn']", "_____no_output_____" ], [ "status = wait(lambda: forecast.describe_forecast_export_job(ForecastExportJobArn = forecast_export_arn_prophet))\nassert status", "_____no_output_____" ] ], [ [ "# Step 8. Cleaning up your Resources<a class=\"anchor\" id=\"cleanup\">", "_____no_output_____" ], [ "Once we have completed the above steps, we can start to cleanup the resources we created. All delete jobs, except for `delete_dataset_group` are asynchronous, so we have added the helpful `wait_till_delete` function. \nResource Limits documented <a href=\"https://docs.aws.amazon.com/forecast/latest/dg/limits.html\">here</a>. ", "_____no_output_____" ] ], [ [ "# Delete forecast export for both algorithms\nwait_till_delete(lambda: forecast.delete_forecast_export_job(ForecastExportJobArn = forecast_export_arn_deep_ar))\nwait_till_delete(lambda: forecast.delete_forecast_export_job(ForecastExportJobArn = forecast_export_arn_prophet))", "_____no_output_____" ], [ "# Delete forecast for both algorithms\nwait_till_delete(lambda: forecast.delete_forecast(ForecastArn = forecast_arn_deep_ar))\nwait_till_delete(lambda: forecast.delete_forecast(ForecastArn = forecast_arn_prophet))", "_____no_output_____" ], [ "# Delete predictor for both algorithms\nwait_till_delete(lambda: forecast.delete_predictor(PredictorArn = predictor_arn_deep_ar))\nwait_till_delete(lambda: forecast.delete_predictor(PredictorArn = predictor_arn_prophet))", "_____no_output_____" ], [ "# Delete the target time series and related time series dataset import jobs\nwait_till_delete(lambda: forecast.delete_dataset_import_job(DatasetImportJobArn=ts_dataset_import_job_arn))\nwait_till_delete(lambda: forecast.delete_dataset_import_job(DatasetImportJobArn=rts_dataset_import_job_arn))", "_____no_output_____" ], [ "# Delete the target time series and related time series datasets\nwait_till_delete(lambda: forecast.delete_dataset(DatasetArn=ts_dataset_arn))\nwait_till_delete(lambda: forecast.delete_dataset(DatasetArn=rts_dataset_arn))", "_____no_output_____" ], [ "# Delete dataset group\nforecast.delete_dataset_group(DatasetGroupArn=dataset_group_arn)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Census aggregation scratchpad\n\nBy [Ben Welsh](https://palewi.re/who-is-ben-welsh/)", "_____no_output_____" ] ], [ [ "import math", "_____no_output_____" ] ], [ [ "### Approximation", "_____no_output_____" ], [ "\n", "_____no_output_____" ] ], [ [ "males_under_5, males_under_5_moe = 10154024, 3778", "_____no_output_____" ], [ "females_under_5, females_under_5_moe = 9712936, 3911", "_____no_output_____" ], [ "total_under_5 = males_under_5 + females_under_5", "_____no_output_____" ], [ "total_under_5", "_____no_output_____" ], [ "total_under_5_moe = math.sqrt(males_under_5_moe**2 + females_under_5_moe**2)", "_____no_output_____" ], [ "total_under_5_moe", "_____no_output_____" ] ], [ [ "", "_____no_output_____" ] ], [ [ "def approximate_margin_of_error(*pairs):\n \"\"\"\n Returns the approximate margin of error after combining all of the provided Census Bureau estimates, taking into account each value's margin of error.\n \n Expects a series of arguments, each a paired list with the estimated value first and the margin of error second.\n \"\"\"\n # According to the Census Bureau, when approximating a sum use only the largest zero estimate margin of error, once\n # https://www.documentcloud.org/documents/6162551-20180418-MOE.html#document/p52\n zeros = [p for p in pairs if p[0] == 0]\n if len(zeros) > 1:\n max_zero_margin = max([p[1] for p in zeros])\n not_zero_margins = [p[1] for p in pairs if p[0] != 0]\n margins = [max_zero_margin] + not_zero_margins\n else:\n margins = [p[1] for p in pairs]\n return math.sqrt(sum([m**2 for m in margins])) ", "_____no_output_____" ], [ "approximate_margin_of_error(\n (males_under_5, males_under_5_moe),\n (females_under_5, females_under_5_moe)\n)", "_____no_output_____" ], [ "approximate_margin_of_error(\n [0, 22],\n [0, 22],\n [0, 29],\n [41, 37]\n)", "_____no_output_____" ] ], [ [ "### Aggregating totals", "_____no_output_____" ] ], [ [ "def total(*pairs):\n \"\"\"\n Returns the combined value of all the provided Census Bureau estimates, along with an approximated margin of error.\n \n Expects a series of arguments, each a paired list with the estimated value first and the margin of error second.\n \"\"\"\n return sum([p[0] for p in pairs]), approximate_margin_of_error(*pairs)", "_____no_output_____" ], [ "total(\n (males_under_5, males_under_5_moe),\n (females_under_5, females_under_5_moe)\n)", "_____no_output_____" ], [ "total(\n [0, 22],\n [0, 22],\n [0, 29],\n [41, 37]\n)", "_____no_output_____" ] ], [ [ "### Aggregating medians", "_____no_output_____" ], [ "\n\n\n", "_____no_output_____" ] ], [ [ "def approximate_median(range_list, design_factor=1.5):\n \"\"\"\n Returns the estimated median from a set of ranged totals.\n\n Useful for generated medians for measures like median household income and median agn when aggregating census geographies.\n\n Expects a list of dictionaries with three keys:\n\n min: The minimum value in the range\n max: The maximum value in the range\n n: The number of people, households or other universe figure in the range\n \"\"\"\n # Sort the list\n range_list.sort(key=lambda x: x['min'])\n\n # For each range calculate its min and max value along the universe's scale\n cumulative_n = 0\n for range_ in range_list:\n range_['n_min'] = cumulative_n\n cumulative_n += range_['n']\n range_['n_max'] = cumulative_n\n\n # What is the total number of observations in the universe?\n n = sum([d['n'] for d in range_list])\n \n # What is the estimated midpoint of the n?\n n_midpoint = n / 2.0\n\n # Now use those to determine which group contains the midpoint.\n try:\n n_midpoint_range = next(d for d in range_list if n_midpoint >= d['n_min'] and n_midpoint <= d['n_max'])\n except StopIteration:\n raise StopIteration(\"The n's midpoint does not fall within a data range.\")\n\n # How many households in the midrange are needed to reach the midpoint?\n n_midrange_gap = n_midpoint - n_midpoint_range['n_min']\n\n # What is the proportion of the group that would be needed to get the midpoint?\n n_midrange_gap_percent = n_midrange_gap / n_midpoint_range['n']\n\n # Apply this proportion to the width of the midrange\n n_midrange_gap_adjusted = (n_midpoint_range['max'] - n_midpoint_range['min']) * n_midrange_gap_percent\n\n # Estimate the median\n estimated_median = n_midpoint_range['min'] + n_midrange_gap_adjusted\n \n # Get the standard error for this dataset\n standard_error = (design_factor * math.sqrt((99/n)*(50**2))) / 100\n \n # Use the standard error to calculate the p values\n p_lower = (.5 - standard_error)\n p_upper = (.5 + standard_error)\n \n # Estimate the p_lower and p_upper n values\n p_lower_n = n * p_lower\n p_upper_n = n * p_upper\n \n # Find the ranges the p values fall within\n try:\n p_lower_range_i, p_lower_range = next(\n (i, d) for i, d in enumerate(range_list)\n if p_lower_n >= d['n_min'] and p_lower_n <= d['n_max']\n )\n except StopIteration:\n raise StopIteration(\"The n's lower p value does not fall within a data range.\")\n\n try:\n p_upper_range_i, p_upper_range = next(\n (i, d) for i, d in enumerate(range_list)\n if p_upper_n >= d['n_min'] and p_upper_n <= d['n_max']\n )\n except StopIteration:\n raise StopIteration(\"The n's higher p value does not fall within a data range.\")\n \n # Use these values to estimate the lower bound of the confidence interval\n p_lower_a1 = p_lower_range['min']\n try:\n p_lower_a2 = range_list[p_lower_range_i+1]['min']\n except IndexError:\n p_lower_a2 = p_lower_range['max']\n p_lower_c1 = p_lower_range['n_min'] / n\n try:\n p_lower_c2 = range_list[p_lower_range_i+1]['n_min'] / n\n except IndexError:\n p_lower_c2 = p_lower_range['n_max'] / n\n lower_bound = ((p_lower - p_lower_c1) / (p_lower_c2 - p_lower_c1)) * (p_lower_a2 - p_lower_a1) + p_lower_a1\n\n # Same for the upper bound\n p_upper_a1 = p_upper_range['min']\n try:\n p_upper_a2 = range_list[p_upper_range_i+1]['min']\n except IndexError:\n p_upper_a2 = p_upper_range['max']\n p_upper_c1 = p_upper_range['n_min'] / n\n try:\n p_upper_c2 = range_list[p_upper_range_i+1]['n_min'] / n\n except IndexError:\n p_upper_c2 = p_upper_range['n_max'] / n\n upper_bound = ((p_upper - p_upper_c1) / (p_upper_c2 - p_upper_c1)) * (p_upper_a2 - p_upper_a1) + p_upper_a1\n \n # Calculate the standard error of the median\n standard_error_median = 0.5 * (upper_bound - lower_bound)\n \n # Calculate the margin of error at the 90% confidence level\n margin_of_error = 1.645 * standard_error_median\n \n # Return the result\n return estimated_median, margin_of_error", "_____no_output_____" ], [ "income = [\n dict(min=-2500, max=9999, n=186),\n dict(min=10000, max=14999, n=78),\n dict(min=15000, max=19999, n=98),\n dict(min=20000, max=24999, n=287),\n dict(min=25000, max=29999, n=142),\n dict(min=30000, max=34999, n=90),\n dict(min=35000, max=39999, n=107),\n dict(min=40000, max=44999, n=104),\n dict(min=45000, max=49999, n=178),\n dict(min=50000, max=59999, n=106),\n dict(min=60000, max=74999, n=177),\n dict(min=75000, max=99999, n=262),\n dict(min=100000, max=124999, n=77),\n dict(min=125000, max=149999, n=100),\n dict(min=150000, max=199999, n=58),\n dict(min=200000, max=250001, n=18)\n]", "_____no_output_____" ], [ "approximate_median(income)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## Install and setup", "_____no_output_____" ], [ "# use conda instead\nmax python version 3.7\n\n!pip install tensorflow", "_____no_output_____" ], [ "# use conda install graphviz instead\n* also rqeuires conda install python-graphviz\n\n!pip install graphviz", "_____no_output_____" ], [ "# must use pip here - no conda package\n!pip install hiddenlayer", "_____no_output_____" ] ], [ [ "import graphviz", "_____no_output_____" ], [ "d= graphviz.Digraph()\nd.edge('hello','world')\nd", "_____no_output_____" ], [ "!conda env list", "# conda environments:\n#\nbase C:\\ProgramData\\Anaconda3\nmyenv * C:\\Users\\Rob.DESKTOP-HBG5EOT\\.conda\\envs\\myenv\ntf37 C:\\Users\\Rob.DESKTOP-HBG5EOT\\.conda\\envs\\tf37\n\n" ] ] ]

[ "markdown", "code" ]

[ [ "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# Copyright 2020 Google LLC\n#\n# Licensed under the Apache License, Version 2.0 (the \"License\");\n# you may not use this file except in compliance with the License.\n# You may obtain a copy of the License at\n#\n# https://www.apache.org/licenses/LICENSE-2.0\n#\n# Unless required by applicable law or agreed to in writing, software\n# distributed under the License is distributed on an \"AS IS\" BASIS,\n# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.", "_____no_output_____" ] ], [ [ "# Vertex client library: AutoML tabular binary classification model for batch prediction\n\n<table align=\"left\">\n <td>\n <a href=\"https://colab.research.google.com/github/GoogleCloudPlatform/vertex-ai-samples/blob/master/notebooks/community/gapic/automl/showcase_automl_tabular_binary_classification_batch.ipynb\">\n <img src=\"https://cloud.google.com/ml-engine/images/colab-logo-32px.png\" alt=\"Colab logo\"> Run in Colab\n </a>\n </td>\n <td>\n <a href=\"https://github.com/GoogleCloudPlatform/vertex-ai-samples/blob/master/notebooks/community/gapic/automl/showcase_automl_tabular_binary_classification_batch.ipynb\">\n <img src=\"https://cloud.google.com/ml-engine/images/github-logo-32px.png\" alt=\"GitHub logo\">\n View on GitHub\n </a>\n </td>\n</table>\n<br/><br/><br/>", "_____no_output_____" ], [ "## Overview\n\n\nThis tutorial demonstrates how to use the Vertex client library for Python to create tabular binary classification models and do batch prediction using Google Cloud's [AutoML](https://cloud.google.com/vertex-ai/docs/start/automl-users).", "_____no_output_____" ], [ "### Dataset\n\nThe dataset used for this tutorial is the [Bank Marketing](gs://cloud-ml-tables-data/bank-marketing.csv). This dataset does not require any feature engineering. The version of the dataset you will use in this tutorial is stored in a public Cloud Storage bucket.", "_____no_output_____" ], [ "### Objective\n\nIn this tutorial, you create an AutoML tabular binary classification model from a Python script, and then do a batch prediction using the Vertex client library. You can alternatively create and deploy models using the `gcloud` command-line tool or online using the Google Cloud Console.\n\nThe steps performed include:\n\n- Create a Vertex `Dataset` resource.\n- Train the model.\n- View the model evaluation.\n- Make a batch prediction.\n\nThere is one key difference between using batch prediction and using online prediction:\n\n* Prediction Service: Does an on-demand prediction for the entire set of instances (i.e., one or more data items) and returns the results in real-time.\n\n* Batch Prediction Service: Does a queued (batch) prediction for the entire set of instances in the background and stores the results in a Cloud Storage bucket when ready.", "_____no_output_____" ], [ "### Costs\n\nThis tutorial uses billable components of Google Cloud (GCP):\n\n* Vertex AI\n* Cloud Storage\n\nLearn about [Vertex AI\npricing](https://cloud.google.com/vertex-ai/pricing) and [Cloud Storage\npricing](https://cloud.google.com/storage/pricing), and use the [Pricing\nCalculator](https://cloud.google.com/products/calculator/)\nto generate a cost estimate based on your projected usage.", "_____no_output_____" ], [ "## Installation\n\nInstall the latest version of Vertex client library.", "_____no_output_____" ] ], [ [ "import os\nimport sys\n\n# Google Cloud Notebook\nif os.path.exists(\"/opt/deeplearning/metadata/env_version\"):\n USER_FLAG = \"--user\"\nelse:\n USER_FLAG = \"\"\n\n! pip3 install -U google-cloud-aiplatform $USER_FLAG", "_____no_output_____" ] ], [ [ "Install the latest GA version of *google-cloud-storage* library as well.", "_____no_output_____" ] ], [ [ "! pip3 install -U google-cloud-storage $USER_FLAG", "_____no_output_____" ] ], [ [ "### Restart the kernel\n\nOnce you've installed the Vertex client library and Google *cloud-storage*, you need to restart the notebook kernel so it can find the packages.", "_____no_output_____" ] ], [ [ "if not os.getenv(\"IS_TESTING\"):\n # Automatically restart kernel after installs\n import IPython\n\n app = IPython.Application.instance()\n app.kernel.do_shutdown(True)", "_____no_output_____" ] ], [ [ "## Before you begin\n\n### GPU runtime\n\n*Make sure you're running this notebook in a GPU runtime if you have that option. In Colab, select* **Runtime > Change Runtime Type > GPU**\n\n### Set up your Google Cloud project\n\n**The following steps are required, regardless of your notebook environment.**\n\n1. [Select or create a Google Cloud project](https://console.cloud.google.com/cloud-resource-manager). When you first create an account, you get a $300 free credit towards your compute/storage costs.\n\n2. [Make sure that billing is enabled for your project.](https://cloud.google.com/billing/docs/how-to/modify-project)\n\n3. [Enable the Vertex APIs and Compute Engine APIs.](https://console.cloud.google.com/flows/enableapi?apiid=ml.googleapis.com,compute_component)\n\n4. [The Google Cloud SDK](https://cloud.google.com/sdk) is already installed in Google Cloud Notebook.\n\n5. Enter your project ID in the cell below. Then run the cell to make sure the\nCloud SDK uses the right project for all the commands in this notebook.\n\n**Note**: Jupyter runs lines prefixed with `!` as shell commands, and it interpolates Python variables prefixed with `$` into these commands.", "_____no_output_____" ] ], [ [ "PROJECT_ID = \"[your-project-id]\" # @param {type:\"string\"}", "_____no_output_____" ], [ "if PROJECT_ID == \"\" or PROJECT_ID is None or PROJECT_ID == \"[your-project-id]\":\n # Get your GCP project id from gcloud\n shell_output = !gcloud config list --format 'value(core.project)' 2>/dev/null\n PROJECT_ID = shell_output[0]\n print(\"Project ID:\", PROJECT_ID)", "_____no_output_____" ], [ "! gcloud config set project $PROJECT_ID", "_____no_output_____" ] ], [ [ "#### Region\n\nYou can also change the `REGION` variable, which is used for operations\nthroughout the rest of this notebook. Below are regions supported for Vertex. We recommend that you choose the region closest to you.\n\n- Americas: `us-central1`\n- Europe: `europe-west4`\n- Asia Pacific: `asia-east1`\n\nYou may not use a multi-regional bucket for training with Vertex. Not all regions provide support for all Vertex services. For the latest support per region, see the [Vertex locations documentation](https://cloud.google.com/vertex-ai/docs/general/locations)", "_____no_output_____" ] ], [ [ "REGION = \"us-central1\" # @param {type: \"string\"}", "_____no_output_____" ] ], [ [ "#### Timestamp\n\nIf you are in a live tutorial session, you might be using a shared test account or project. To avoid name collisions between users on resources created, you create a timestamp for each instance session, and append onto the name of resources which will be created in this tutorial.", "_____no_output_____" ] ], [ [ "from datetime import datetime\n\nTIMESTAMP = datetime.now().strftime(\"%Y%m%d%H%M%S\")", "_____no_output_____" ] ], [ [ "### Authenticate your Google Cloud account\n\n**If you are using Google Cloud Notebook**, your environment is already authenticated. Skip this step.\n\n**If you are using Colab**, run the cell below and follow the instructions when prompted to authenticate your account via oAuth.\n\n**Otherwise**, follow these steps:\n\nIn the Cloud Console, go to the [Create service account key](https://console.cloud.google.com/apis/credentials/serviceaccountkey) page.\n\n**Click Create service account**.\n\nIn the **Service account name** field, enter a name, and click **Create**.\n\nIn the **Grant this service account access to project** section, click the Role drop-down list. Type \"Vertex\" into the filter box, and select **Vertex Administrator**. Type \"Storage Object Admin\" into the filter box, and select **Storage Object Admin**.\n\nClick Create. A JSON file that contains your key downloads to your local environment.\n\nEnter the path to your service account key as the GOOGLE_APPLICATION_CREDENTIALS variable in the cell below and run the cell.", "_____no_output_____" ] ], [ [ "# If you are running this notebook in Colab, run this cell and follow the\n# instructions to authenticate your GCP account. This provides access to your\n# Cloud Storage bucket and lets you submit training jobs and prediction\n# requests.\n\n# If on Google Cloud Notebook, then don't execute this code\nif not os.path.exists(\"/opt/deeplearning/metadata/env_version\"):\n if \"google.colab\" in sys.modules:\n from google.colab import auth as google_auth\n\n google_auth.authenticate_user()\n\n # If you are running this notebook locally, replace the string below with the\n # path to your service account key and run this cell to authenticate your GCP\n # account.\n elif not os.getenv(\"IS_TESTING\"):\n %env GOOGLE_APPLICATION_CREDENTIALS ''", "_____no_output_____" ] ], [ [ "### Create a Cloud Storage bucket\n\n**The following steps are required, regardless of your notebook environment.**\n\nThis tutorial is designed to use training data that is in a public Cloud Storage bucket and a local Cloud Storage bucket for your batch predictions. You may alternatively use your own training data that you have stored in a local Cloud Storage bucket.\n\nSet the name of your Cloud Storage bucket below. Bucket names must be globally unique across all Google Cloud projects, including those outside of your organization.", "_____no_output_____" ] ], [ [ "BUCKET_NAME = \"gs://[your-bucket-name]\" # @param {type:\"string\"}", "_____no_output_____" ], [ "if BUCKET_NAME == \"\" or BUCKET_NAME is None or BUCKET_NAME == \"gs://[your-bucket-name]\":\n BUCKET_NAME = \"gs://\" + PROJECT_ID + \"aip-\" + TIMESTAMP", "_____no_output_____" ] ], [ [ "**Only if your bucket doesn't already exist**: Run the following cell to create your Cloud Storage bucket.", "_____no_output_____" ] ], [ [ "! gsutil mb -l $REGION $BUCKET_NAME", "_____no_output_____" ] ], [ [ "Finally, validate access to your Cloud Storage bucket by examining its contents:", "_____no_output_____" ] ], [ [ "! gsutil ls -al $BUCKET_NAME", "_____no_output_____" ] ], [ [ "### Set up variables\n\nNext, set up some variables used throughout the tutorial.\n### Import libraries and define constants", "_____no_output_____" ], [ "#### Import Vertex client library\n\nImport the Vertex client library into our Python environment.", "_____no_output_____" ] ], [ [ "import time\n\nfrom google.cloud.aiplatform import gapic as aip\nfrom google.protobuf import json_format\nfrom google.protobuf.struct_pb2 import Struct, Value", "_____no_output_____" ] ], [ [ "#### Vertex constants\n\nSetup up the following constants for Vertex:\n\n- `API_ENDPOINT`: The Vertex API service endpoint for dataset, model, job, pipeline and endpoint services.\n- `PARENT`: The Vertex location root path for dataset, model, job, pipeline and endpoint resources.", "_____no_output_____" ] ], [ [ "# API service endpoint\nAPI_ENDPOINT = \"{}-aiplatform.googleapis.com\".format(REGION)\n\n# Vertex location root path for your dataset, model and endpoint resources\nPARENT = \"projects/\" + PROJECT_ID + \"/locations/\" + REGION", "_____no_output_____" ] ], [ [ "#### AutoML constants\n\nSet constants unique to AutoML datasets and training:\n\n- Dataset Schemas: Tells the `Dataset` resource service which type of dataset it is.\n- Data Labeling (Annotations) Schemas: Tells the `Dataset` resource service how the data is labeled (annotated).\n- Dataset Training Schemas: Tells the `Pipeline` resource service the task (e.g., classification) to train the model for.", "_____no_output_____" ] ], [ [ "# Tabular Dataset type\nDATA_SCHEMA = \"gs://google-cloud-aiplatform/schema/dataset/metadata/tables_1.0.0.yaml\"\n# Tabular Labeling type\nLABEL_SCHEMA = (\n \"gs://google-cloud-aiplatform/schema/dataset/ioformat/table_io_format_1.0.0.yaml\"\n)\n# Tabular Training task\nTRAINING_SCHEMA = \"gs://google-cloud-aiplatform/schema/trainingjob/definition/automl_tables_1.0.0.yaml\"", "_____no_output_____" ] ], [ [ "#### Hardware Accelerators\n\nSet the hardware accelerators (e.g., GPU), if any, for prediction.\n\nSet the variable `DEPLOY_GPU/DEPLOY_NGPU` to use a container image supporting a GPU and the number of GPUs allocated to the virtual machine (VM) instance. For example, to use a GPU container image with 4 Nvidia Telsa K80 GPUs allocated to each VM, you would specify:\n\n (aip.AcceleratorType.NVIDIA_TESLA_K80, 4)\n\nFor GPU, available accelerators include:\n - aip.AcceleratorType.NVIDIA_TESLA_K80\n - aip.AcceleratorType.NVIDIA_TESLA_P100\n - aip.AcceleratorType.NVIDIA_TESLA_P4\n - aip.AcceleratorType.NVIDIA_TESLA_T4\n - aip.AcceleratorType.NVIDIA_TESLA_V100\n\nOtherwise specify `(None, None)` to use a container image to run on a CPU.", "_____no_output_____" ] ], [ [ "if os.getenv(\"IS_TESTING_DEPOLY_GPU\"):\n DEPLOY_GPU, DEPLOY_NGPU = (\n aip.AcceleratorType.NVIDIA_TESLA_K80,\n int(os.getenv(\"IS_TESTING_DEPOLY_GPU\")),\n )\nelse:\n DEPLOY_GPU, DEPLOY_NGPU = (aip.AcceleratorType.NVIDIA_TESLA_K80, 1)", "_____no_output_____" ] ], [ [ "#### Container (Docker) image\n\nFor AutoML batch prediction, the container image for the serving binary is pre-determined by the Vertex prediction service. More specifically, the service will pick the appropriate container for the model depending on the hardware accelerator you selected.", "_____no_output_____" ], [ "#### Machine Type\n\nNext, set the machine type to use for prediction.\n\n- Set the variable `DEPLOY_COMPUTE` to configure the compute resources for the VM you will use for prediction.\n - `machine type`\n - `n1-standard`: 3.75GB of memory per vCPU.\n - `n1-highmem`: 6.5GB of memory per vCPU\n - `n1-highcpu`: 0.9 GB of memory per vCPU\n - `vCPUs`: number of \\[2, 4, 8, 16, 32, 64, 96 \\]\n\n*Note: You may also use n2 and e2 machine types for training and deployment, but they do not support GPUs*", "_____no_output_____" ] ], [ [ "if os.getenv(\"IS_TESTING_DEPLOY_MACHINE\"):\n MACHINE_TYPE = os.getenv(\"IS_TESTING_DEPLOY_MACHINE\")\nelse:\n MACHINE_TYPE = \"n1-standard\"\n\nVCPU = \"4\"\nDEPLOY_COMPUTE = MACHINE_TYPE + \"-\" + VCPU\nprint(\"Deploy machine type\", DEPLOY_COMPUTE)", "_____no_output_____" ] ], [ [ "# Tutorial\n\nNow you are ready to start creating your own AutoML tabular binary classification model.", "_____no_output_____" ], [ "## Set up clients\n\nThe Vertex client library works as a client/server model. On your side (the Python script) you will create a client that sends requests and receives responses from the Vertex server.\n\nYou will use different clients in this tutorial for different steps in the workflow. So set them all up upfront.\n\n- Dataset Service for `Dataset` resources.\n- Model Service for `Model` resources.\n- Pipeline Service for training.\n- Job Service for batch prediction and custom training.", "_____no_output_____" ] ], [ [ "# client options same for all services\nclient_options = {\"api_endpoint\": API_ENDPOINT}\n\n\ndef create_dataset_client():\n client = aip.DatasetServiceClient(client_options=client_options)\n return client\n\n\ndef create_model_client():\n client = aip.ModelServiceClient(client_options=client_options)\n return client\n\n\ndef create_pipeline_client():\n client = aip.PipelineServiceClient(client_options=client_options)\n return client\n\n\ndef create_job_client():\n client = aip.JobServiceClient(client_options=client_options)\n return client\n\n\nclients = {}\nclients[\"dataset\"] = create_dataset_client()\nclients[\"model\"] = create_model_client()\nclients[\"pipeline\"] = create_pipeline_client()\nclients[\"job\"] = create_job_client()\n\nfor client in clients.items():\n print(client)", "_____no_output_____" ] ], [ [ "## Dataset\n\nNow that your clients are ready, your first step is to create a `Dataset` resource instance. This step differs from Vision, Video and Language. For those products, after the `Dataset` resource is created, one then separately imports the data, using the `import_data` method.\n\nFor tabular, importing of the data is deferred until the training pipeline starts training the model. What do we do different? Well, first you won't be calling the `import_data` method. Instead, when you create the dataset instance you specify the Cloud Storage location of the CSV file or BigQuery location of the data table, which contains your tabular data as part of the `Dataset` resource's metadata.\n\n#### Cloud Storage\n\n`metadata = {\"input_config\": {\"gcs_source\": {\"uri\": [gcs_uri]}}}`\n\nThe format for a Cloud Storage path is:\n\n gs://[bucket_name]/[folder(s)/[file]\n\n#### BigQuery\n\n`metadata = {\"input_config\": {\"bigquery_source\": {\"uri\": [gcs_uri]}}}`\n\nThe format for a BigQuery path is:\n\n bq://[collection].[dataset].[table]\n\nNote that the `uri` field is a list, whereby you can input multiple CSV files or BigQuery tables when your data is split across files.", "_____no_output_____" ], [ "### Data preparation\n\nThe Vertex `Dataset` resource for tabular has a couple of requirements for your tabular data.\n\n- Must be in a CSV file or a BigQuery query.", "_____no_output_____" ], [ "#### CSV\n\nFor tabular binary classification, the CSV file has a few requirements:\n\n- The first row must be the heading -- note how this is different from Vision, Video and Language where the requirement is no heading.\n- All but one column are features.\n- One column is the label, which you will specify when you subsequently create the training pipeline.", "_____no_output_____" ], [ "#### Location of Cloud Storage training data.\n\nNow set the variable `IMPORT_FILE` to the location of the CSV index file in Cloud Storage.", "_____no_output_____" ] ], [ [ "IMPORT_FILE = \"gs://cloud-ml-tables-data/bank-marketing.csv\"", "_____no_output_____" ] ], [ [ "#### Quick peek at your data\n\nYou will use a version of the Bank Marketing dataset that is stored in a public Cloud Storage bucket, using a CSV index file.\n\nStart by doing a quick peek at the data. You count the number of examples by counting the number of rows in the CSV index file (`wc -l`) and then peek at the first few rows.\n\nYou also need for training to know the heading name of the label column, which is save as `label_column`. For this dataset, it is the last column in the CSV file.", "_____no_output_____" ] ], [ [ "count = ! gsutil cat $IMPORT_FILE | wc -l\nprint(\"Number of Examples\", int(count[0]))\n\nprint(\"First 10 rows\")\n! gsutil cat $IMPORT_FILE | head\n\nheading = ! gsutil cat $IMPORT_FILE | head -n1\nlabel_column = str(heading).split(\",\")[-1].split(\"'\")[0]\nprint(\"Label Column Name\", label_column)\nif label_column is None:\n raise Exception(\"label column missing\")", "_____no_output_____" ] ], [ [ "## Dataset\n\nNow that your clients are ready, your first step in training a model is to create a managed dataset instance, and then upload your labeled data to it.\n\n### Create `Dataset` resource instance\n\nUse the helper function `create_dataset` to create the instance of a `Dataset` resource. This function does the following:\n\n1. Uses the dataset client service.\n2. Creates an Vertex `Dataset` resource (`aip.Dataset`), with the following parameters:\n - `display_name`: The human-readable name you choose to give it.\n - `metadata_schema_uri`: The schema for the dataset type.\n - `metadata`: The Cloud Storage or BigQuery location of the tabular data.\n3. Calls the client dataset service method `create_dataset`, with the following parameters:\n - `parent`: The Vertex location root path for your `Database`, `Model` and `Endpoint` resources.\n - `dataset`: The Vertex dataset object instance you created.\n4. The method returns an `operation` object.\n\nAn `operation` object is how Vertex handles asynchronous calls for long running operations. While this step usually goes fast, when you first use it in your project, there is a longer delay due to provisioning.\n\nYou can use the `operation` object to get status on the operation (e.g., create `Dataset` resource) or to cancel the operation, by invoking an operation method:\n\n| Method | Description |\n| ----------- | ----------- |\n| result() | Waits for the operation to complete and returns a result object in JSON format. |\n| running() | Returns True/False on whether the operation is still running. |\n| done() | Returns True/False on whether the operation is completed. |\n| canceled() | Returns True/False on whether the operation was canceled. |\n| cancel() | Cancels the operation (this may take up to 30 seconds). |", "_____no_output_____" ] ], [ [ "TIMEOUT = 90\n\n\ndef create_dataset(name, schema, src_uri=None, labels=None, timeout=TIMEOUT):\n start_time = time.time()\n try:\n if src_uri.startswith(\"gs://\"):\n metadata = {\"input_config\": {\"gcs_source\": {\"uri\": [src_uri]}}}\n elif src_uri.startswith(\"bq://\"):\n metadata = {\"input_config\": {\"bigquery_source\": {\"uri\": [src_uri]}}}\n dataset = aip.Dataset(\n display_name=name,\n metadata_schema_uri=schema,\n labels=labels,\n metadata=json_format.ParseDict(metadata, Value()),\n )\n\n operation = clients[\"dataset\"].create_dataset(parent=PARENT, dataset=dataset)\n print(\"Long running operation:\", operation.operation.name)\n result = operation.result(timeout=TIMEOUT)\n print(\"time:\", time.time() - start_time)\n print(\"response\")\n print(\" name:\", result.name)\n print(\" display_name:\", result.display_name)\n print(\" metadata_schema_uri:\", result.metadata_schema_uri)\n print(\" metadata:\", dict(result.metadata))\n print(\" create_time:\", result.create_time)\n print(\" update_time:\", result.update_time)\n print(\" etag:\", result.etag)\n print(\" labels:\", dict(result.labels))\n return result\n except Exception as e:\n print(\"exception:\", e)\n return None\n\n\nresult = create_dataset(\"bank-\" + TIMESTAMP, DATA_SCHEMA, src_uri=IMPORT_FILE)", "_____no_output_____" ] ], [ [ "Now save the unique dataset identifier for the `Dataset` resource instance you created.", "_____no_output_____" ] ], [ [ "# The full unique ID for the dataset\ndataset_id = result.name\n# The short numeric ID for the dataset\ndataset_short_id = dataset_id.split(\"/\")[-1]\n\nprint(dataset_id)", "_____no_output_____" ] ], [ [ "## Train the model\n\nNow train an AutoML tabular binary classification model using your Vertex `Dataset` resource. To train the model, do the following steps:\n\n1. Create an Vertex training pipeline for the `Dataset` resource.\n2. Execute the pipeline to start the training.", "_____no_output_____" ], [ "### Create a training pipeline\n\nYou may ask, what do we use a pipeline for? You typically use pipelines when the job (such as training) has multiple steps, generally in sequential order: do step A, do step B, etc. By putting the steps into a pipeline, we gain the benefits of:\n\n1. Being reusable for subsequent training jobs.\n2. Can be containerized and ran as a batch job.\n3. Can be distributed.\n4. All the steps are associated with the same pipeline job for tracking progress.\n\nUse this helper function `create_pipeline`, which takes the following parameters:\n\n- `pipeline_name`: A human readable name for the pipeline job.\n- `model_name`: A human readable name for the model.\n- `dataset`: The Vertex fully qualified dataset identifier.\n- `schema`: The dataset labeling (annotation) training schema.\n- `task`: A dictionary describing the requirements for the training job.\n\nThe helper function calls the `Pipeline` client service'smethod `create_pipeline`, which takes the following parameters:\n\n- `parent`: The Vertex location root path for your `Dataset`, `Model` and `Endpoint` resources.\n- `training_pipeline`: the full specification for the pipeline training job.\n\nLet's look now deeper into the *minimal* requirements for constructing a `training_pipeline` specification:\n\n- `display_name`: A human readable name for the pipeline job.\n- `training_task_definition`: The dataset labeling (annotation) training schema.\n- `training_task_inputs`: A dictionary describing the requirements for the training job.\n- `model_to_upload`: A human readable name for the model.\n- `input_data_config`: The dataset specification.\n - `dataset_id`: The Vertex dataset identifier only (non-fully qualified) -- this is the last part of the fully-qualified identifier.\n - `fraction_split`: If specified, the percentages of the dataset to use for training, test and validation. Otherwise, the percentages are automatically selected by AutoML.", "_____no_output_____" ] ], [ [ "def create_pipeline(pipeline_name, model_name, dataset, schema, task):\n\n dataset_id = dataset.split(\"/\")[-1]\n\n input_config = {\n \"dataset_id\": dataset_id,\n \"fraction_split\": {\n \"training_fraction\": 0.8,\n \"validation_fraction\": 0.1,\n \"test_fraction\": 0.1,\n },\n }\n\n training_pipeline = {\n \"display_name\": pipeline_name,\n \"training_task_definition\": schema,\n \"training_task_inputs\": task,\n \"input_data_config\": input_config,\n \"model_to_upload\": {\"display_name\": model_name},\n }\n\n try:\n pipeline = clients[\"pipeline\"].create_training_pipeline(\n parent=PARENT, training_pipeline=training_pipeline\n )\n print(pipeline)\n except Exception as e:\n print(\"exception:\", e)\n return None\n return pipeline", "_____no_output_____" ] ], [ [ "### Construct the task requirements\n\nNext, construct the task requirements. Unlike other parameters which take a Python (JSON-like) dictionary, the `task` field takes a Google protobuf Struct, which is very similar to a Python dictionary. Use the `json_format.ParseDict` method for the conversion.\n\nThe minimal fields you need to specify are:\n\n- `prediction_type`: Whether we are doing \"classification\" or \"regression\".\n- `target_column`: The CSV heading column name for the column we want to predict (i.e., the label).\n- `train_budget_milli_node_hours`: The maximum time to budget (billed) for training the model, where 1000 = 1 hour.\n- `disable_early_stopping`: Whether True/False to let AutoML use its judgement to stop training early or train for the entire budget.\n- `transformations`: Specifies the feature engineering for each feature column.\n\nFor `transformations`, the list must have an entry for each column. The outer key field indicates the type of feature engineering for the corresponding column. In this tutorial, you set it to `\"auto\"` to tell AutoML to automatically determine it.\n\nFinally, create the pipeline by calling the helper function `create_pipeline`, which returns an instance of a training pipeline object.", "_____no_output_____" ] ], [ [ "TRANSFORMATIONS = [\n {\"auto\": {\"column_name\": \"Age\"}},\n {\"auto\": {\"column_name\": \"Job\"}},\n {\"auto\": {\"column_name\": \"MaritalStatus\"}},\n {\"auto\": {\"column_name\": \"Education\"}},\n {\"auto\": {\"column_name\": \"Default\"}},\n {\"auto\": {\"column_name\": \"Balance\"}},\n {\"auto\": {\"column_name\": \"Housing\"}},\n {\"auto\": {\"column_name\": \"Loan\"}},\n {\"auto\": {\"column_name\": \"Contact\"}},\n {\"auto\": {\"column_name\": \"Day\"}},\n {\"auto\": {\"column_name\": \"Month\"}},\n {\"auto\": {\"column_name\": \"Duration\"}},\n {\"auto\": {\"column_name\": \"Campaign\"}},\n {\"auto\": {\"column_name\": \"PDays\"}},\n {\"auto\": {\"column_name\": \"POutcome\"}},\n]", "_____no_output_____" ], [ "PIPE_NAME = \"bank_pipe-\" + TIMESTAMP\nMODEL_NAME = \"bank_model-\" + TIMESTAMP\n\ntask = Value(\n struct_value=Struct(\n fields={\n \"target_column\": Value(string_value=label_column),\n \"prediction_type\": Value(string_value=\"classification\"),\n \"train_budget_milli_node_hours\": Value(number_value=1000),\n \"disable_early_stopping\": Value(bool_value=False),\n \"transformations\": json_format.ParseDict(TRANSFORMATIONS, Value()),\n }\n )\n)\n\nresponse = create_pipeline(PIPE_NAME, MODEL_NAME, dataset_id, TRAINING_SCHEMA, task)", "_____no_output_____" ] ], [ [ "Now save the unique identifier of the training pipeline you created.", "_____no_output_____" ] ], [ [ "# The full unique ID for the pipeline\npipeline_id = response.name\n# The short numeric ID for the pipeline\npipeline_short_id = pipeline_id.split(\"/\")[-1]\n\nprint(pipeline_id)", "_____no_output_____" ] ], [ [ "### Get information on a training pipeline\n\nNow get pipeline information for just this training pipeline instance. The helper function gets the job information for just this job by calling the the job client service's `get_training_pipeline` method, with the following parameter:\n\n- `name`: The Vertex fully qualified pipeline identifier.\n\nWhen the model is done training, the pipeline state will be `PIPELINE_STATE_SUCCEEDED`.", "_____no_output_____" ] ], [ [ "def get_training_pipeline(name, silent=False):\n response = clients[\"pipeline\"].get_training_pipeline(name=name)\n if silent:\n return response\n\n print(\"pipeline\")\n print(\" name:\", response.name)\n print(\" display_name:\", response.display_name)\n print(\" state:\", response.state)\n print(\" training_task_definition:\", response.training_task_definition)\n print(\" training_task_inputs:\", dict(response.training_task_inputs))\n print(\" create_time:\", response.create_time)\n print(\" start_time:\", response.start_time)\n print(\" end_time:\", response.end_time)\n print(\" update_time:\", response.update_time)\n print(\" labels:\", dict(response.labels))\n return response\n\n\nresponse = get_training_pipeline(pipeline_id)", "_____no_output_____" ] ], [ [ "# Deployment\n\nTraining the above model may take upwards of 30 minutes time.\n\nOnce your model is done training, you can calculate the actual time it took to train the model by subtracting `end_time` from `start_time`. For your model, you will need to know the fully qualified Vertex Model resource identifier, which the pipeline service assigned to it. You can get this from the returned pipeline instance as the field `model_to_deploy.name`.", "_____no_output_____" ] ], [ [ "while True:\n response = get_training_pipeline(pipeline_id, True)\n if response.state != aip.PipelineState.PIPELINE_STATE_SUCCEEDED:\n print(\"Training job has not completed:\", response.state)\n model_to_deploy_id = None\n if response.state == aip.PipelineState.PIPELINE_STATE_FAILED:\n raise Exception(\"Training Job Failed\")\n else:\n model_to_deploy = response.model_to_upload\n model_to_deploy_id = model_to_deploy.name\n print(\"Training Time:\", response.end_time - response.start_time)\n break\n time.sleep(60)\n\nprint(\"model to deploy:\", model_to_deploy_id)", "_____no_output_____" ] ], [ [ "## Model information\n\nNow that your model is trained, you can get some information on your model.", "_____no_output_____" ], [ "## Evaluate the Model resource\n\nNow find out how good the model service believes your model is. As part of training, some portion of the dataset was set aside as the test (holdout) data, which is used by the pipeline service to evaluate the model.", "_____no_output_____" ], [ "### List evaluations for all slices\n\nUse this helper function `list_model_evaluations`, which takes the following parameter:\n\n- `name`: The Vertex fully qualified model identifier for the `Model` resource.\n\nThis helper function uses the model client service's `list_model_evaluations` method, which takes the same parameter. The response object from the call is a list, where each element is an evaluation metric.\n\nFor each evaluation (you probably only have one) we then print all the key names for each metric in the evaluation, and for a small set (`logLoss` and `auPrc`) you will print the result.", "_____no_output_____" ] ], [ [ "def list_model_evaluations(name):\n response = clients[\"model\"].list_model_evaluations(parent=name)\n for evaluation in response:\n print(\"model_evaluation\")\n print(\" name:\", evaluation.name)\n print(\" metrics_schema_uri:\", evaluation.metrics_schema_uri)\n metrics = json_format.MessageToDict(evaluation._pb.metrics)\n for metric in metrics.keys():\n print(metric)\n print(\"logloss\", metrics[\"logLoss\"])\n print(\"auPrc\", metrics[\"auPrc\"])\n\n return evaluation.name\n\n\nlast_evaluation = list_model_evaluations(model_to_deploy_id)", "_____no_output_____" ] ], [ [ "## Model deployment for batch prediction\n\nNow deploy the trained Vertex `Model` resource you created for batch prediction. This differs from deploying a `Model` resource for on-demand prediction.\n\nFor online prediction, you:\n\n1. Create an `Endpoint` resource for deploying the `Model` resource to.\n\n2. Deploy the `Model` resource to the `Endpoint` resource.\n\n3. Make online prediction requests to the `Endpoint` resource.\n\nFor batch-prediction, you:\n\n1. Create a batch prediction job.\n\n2. The job service will provision resources for the batch prediction request.\n\n3. The results of the batch prediction request are returned to the caller.\n\n4. The job service will unprovision the resoures for the batch prediction request.", "_____no_output_____" ], [ "## Make a batch prediction request\n\nNow do a batch prediction to your deployed model.", "_____no_output_____" ], [ "### Make test items\n\nYou will use synthetic data as a test data items. Don't be concerned that we are using synthetic data -- we just want to demonstrate how to make a prediction.", "_____no_output_____" ] ], [ [ "HEADING = \"Age,Job,MaritalStatus,Education,Default,Balance,Housing,Loan,Contact,Day,Month,Duration,Campaign,PDays,Previous,POutcome,Deposit\"\nINSTANCE_1 = (\n \"58,managment,married,teritary,no,2143,yes,no,unknown,5,may,261,1,-1,0, unknown\"\n)\nINSTANCE_2 = (\n \"44,technician,single,secondary,no,39,yes,no,unknown,5,may,151,1,-1,0,unknown\"\n)", "_____no_output_____" ] ], [ [ "### Make the batch input file\n\nNow make a batch input file, which you will store in your local Cloud Storage bucket. Unlike image, video and text, the batch input file for tabular is only supported for CSV. For CSV file, you make:\n\n- The first line is the heading with the feature (fields) heading names.\n- Each remaining line is a separate prediction request with the corresponding feature values.\n\nFor example:\n\n \"feature_1\", \"feature_2\". ...\n value_1, value_2, ...", "_____no_output_____" ] ], [ [ "import tensorflow as tf\n\ngcs_input_uri = BUCKET_NAME + \"/test.csv\"\nwith tf.io.gfile.GFile(gcs_input_uri, \"w\") as f:\n f.write(HEADING + \"\\n\")\n f.write(str(INSTANCE_1) + \"\\n\")\n f.write(str(INSTANCE_2) + \"\\n\")\n\nprint(gcs_input_uri)\n! gsutil cat $gcs_input_uri", "_____no_output_____" ] ], [ [ "### Compute instance scaling\n\nYou have several choices on scaling the compute instances for handling your batch prediction requests:\n\n- Single Instance: The batch prediction requests are processed on a single compute instance.\n - Set the minimum (`MIN_NODES`) and maximum (`MAX_NODES`) number of compute instances to one.\n\n- Manual Scaling: The batch prediction requests are split across a fixed number of compute instances that you manually specified.\n - Set the minimum (`MIN_NODES`) and maximum (`MAX_NODES`) number of compute instances to the same number of nodes. When a model is first deployed to the instance, the fixed number of compute instances are provisioned and batch prediction requests are evenly distributed across them.\n\n- Auto Scaling: The batch prediction requests are split across a scaleable number of compute instances.\n - Set the minimum (`MIN_NODES`) number of compute instances to provision when a model is first deployed and to de-provision, and set the maximum (`MAX_NODES) number of compute instances to provision, depending on load conditions.\n\nThe minimum number of compute instances corresponds to the field `min_replica_count` and the maximum number of compute instances corresponds to the field `max_replica_count`, in your subsequent deployment request.", "_____no_output_____" ] ], [ [ "MIN_NODES = 1\nMAX_NODES = 1", "_____no_output_____" ] ], [ [ "### Make batch prediction request\n\nNow that your batch of two test items is ready, let's do the batch request. Use this helper function `create_batch_prediction_job`, with the following parameters:\n\n- `display_name`: The human readable name for the prediction job.\n- `model_name`: The Vertex fully qualified identifier for the `Model` resource.\n- `gcs_source_uri`: The Cloud Storage path to the input file -- which you created above.\n- `gcs_destination_output_uri_prefix`: The Cloud Storage path that the service will write the predictions to.\n- `parameters`: Additional filtering parameters for serving prediction results.\n\nThe helper function calls the job client service's `create_batch_prediction_job` metho, with the following parameters:\n\n- `parent`: The Vertex location root path for Dataset, Model and Pipeline resources.\n- `batch_prediction_job`: The specification for the batch prediction job.\n\nLet's now dive into the specification for the `batch_prediction_job`:\n\n- `display_name`: The human readable name for the prediction batch job.\n- `model`: The Vertex fully qualified identifier for the `Model` resource.\n- `dedicated_resources`: The compute resources to provision for the batch prediction job.\n - `machine_spec`: The compute instance to provision. Use the variable you set earlier `DEPLOY_GPU != None` to use a GPU; otherwise only a CPU is allocated.\n - `starting_replica_count`: The number of compute instances to initially provision, which you set earlier as the variable `MIN_NODES`.\n - `max_replica_count`: The maximum number of compute instances to scale to, which you set earlier as the variable `MAX_NODES`.\n- `model_parameters`: Additional filtering parameters for serving prediction results. *Note*, image segmentation models do not support additional parameters.\n- `input_config`: The input source and format type for the instances to predict.\n - `instances_format`: The format of the batch prediction request file: `csv` only supported.\n - `gcs_source`: A list of one or more Cloud Storage paths to your batch prediction requests.\n- `output_config`: The output destination and format for the predictions.\n - `prediction_format`: The format of the batch prediction response file: `csv` only supported.\n - `gcs_destination`: The output destination for the predictions.\n\nThis call is an asychronous operation. You will print from the response object a few select fields, including:\n\n- `name`: The Vertex fully qualified identifier assigned to the batch prediction job.\n- `display_name`: The human readable name for the prediction batch job.\n- `model`: The Vertex fully qualified identifier for the Model resource.\n- `generate_explanations`: Whether True/False explanations were provided with the predictions (explainability).\n- `state`: The state of the prediction job (pending, running, etc).\n\nSince this call will take a few moments to execute, you will likely get `JobState.JOB_STATE_PENDING` for `state`.", "_____no_output_____" ] ], [ [ "BATCH_MODEL = \"bank_batch-\" + TIMESTAMP\n\n\ndef create_batch_prediction_job(\n display_name,\n model_name,\n gcs_source_uri,\n gcs_destination_output_uri_prefix,\n parameters=None,\n):\n\n if DEPLOY_GPU:\n machine_spec = {\n \"machine_type\": DEPLOY_COMPUTE,\n \"accelerator_type\": DEPLOY_GPU,\n \"accelerator_count\": DEPLOY_NGPU,\n }\n else:\n machine_spec = {\n \"machine_type\": DEPLOY_COMPUTE,\n \"accelerator_count\": 0,\n }\n\n batch_prediction_job = {\n \"display_name\": display_name,\n # Format: 'projects/{project}/locations/{location}/models/{model_id}'\n \"model\": model_name,\n \"model_parameters\": json_format.ParseDict(parameters, Value()),\n \"input_config\": {\n \"instances_format\": IN_FORMAT,\n \"gcs_source\": {\"uris\": [gcs_source_uri]},\n },\n \"output_config\": {\n \"predictions_format\": OUT_FORMAT,\n \"gcs_destination\": {\"output_uri_prefix\": gcs_destination_output_uri_prefix},\n },\n \"dedicated_resources\": {\n \"machine_spec\": machine_spec,\n \"starting_replica_count\": MIN_NODES,\n \"max_replica_count\": MAX_NODES,\n },\n }\n response = clients[\"job\"].create_batch_prediction_job(\n parent=PARENT, batch_prediction_job=batch_prediction_job\n )\n print(\"response\")\n print(\" name:\", response.name)\n print(\" display_name:\", response.display_name)\n print(\" model:\", response.model)\n try:\n print(\" generate_explanation:\", response.generate_explanation)\n except:\n pass\n print(\" state:\", response.state)\n print(\" create_time:\", response.create_time)\n print(\" start_time:\", response.start_time)\n print(\" end_time:\", response.end_time)\n print(\" update_time:\", response.update_time)\n print(\" labels:\", response.labels)\n return response\n\n\nIN_FORMAT = \"csv\"\nOUT_FORMAT = \"csv\" # [csv]\n\nresponse = create_batch_prediction_job(\n BATCH_MODEL, model_to_deploy_id, gcs_input_uri, BUCKET_NAME, None\n)", "_____no_output_____" ] ], [ [ "Now get the unique identifier for the batch prediction job you created.", "_____no_output_____" ] ], [ [ "# The full unique ID for the batch job\nbatch_job_id = response.name\n# The short numeric ID for the batch job\nbatch_job_short_id = batch_job_id.split(\"/\")[-1]\n\nprint(batch_job_id)", "_____no_output_____" ] ], [ [ "### Get information on a batch prediction job\n\nUse this helper function `get_batch_prediction_job`, with the following paramter:\n\n- `job_name`: The Vertex fully qualified identifier for the batch prediction job.\n\nThe helper function calls the job client service's `get_batch_prediction_job` method, with the following paramter:\n\n- `name`: The Vertex fully qualified identifier for the batch prediction job. In this tutorial, you will pass it the Vertex fully qualified identifier for your batch prediction job -- `batch_job_id`\n\nThe helper function will return the Cloud Storage path to where the predictions are stored -- `gcs_destination`.", "_____no_output_____" ] ], [ [ "def get_batch_prediction_job(job_name, silent=False):\n response = clients[\"job\"].get_batch_prediction_job(name=job_name)\n if silent:\n return response.output_config.gcs_destination.output_uri_prefix, response.state\n\n print(\"response\")\n print(\" name:\", response.name)\n print(\" display_name:\", response.display_name)\n print(\" model:\", response.model)\n try: # not all data types support explanations\n print(\" generate_explanation:\", response.generate_explanation)\n except:\n pass\n print(\" state:\", response.state)\n print(\" error:\", response.error)\n gcs_destination = response.output_config.gcs_destination\n print(\" gcs_destination\")\n print(\" output_uri_prefix:\", gcs_destination.output_uri_prefix)\n return gcs_destination.output_uri_prefix, response.state\n\n\npredictions, state = get_batch_prediction_job(batch_job_id)", "_____no_output_____" ] ], [ [ "### Get Predictions\n\nWhen the batch prediction is done processing, the job state will be `JOB_STATE_SUCCEEDED`.\n\nFinally you view the predictions stored at the Cloud Storage path you set as output. The predictions will be in a CSV format, which you indicated at the time you made the batch prediction job, under a subfolder starting with the name `prediction`, and under that folder will be a file called `predictions*.csv`.\n\nNow display (cat) the contents. You will see multiple rows, one for each prediction.\n\nFor each prediction:\n\n- The first four fields are the values (features) you did the prediction on.\n- The remaining fields are the confidence values, between 0 and 1, for each prediction.", "_____no_output_____" ] ], [ [ "def get_latest_predictions(gcs_out_dir):\n \"\"\" Get the latest prediction subfolder using the timestamp in the subfolder name\"\"\"\n folders = !gsutil ls $gcs_out_dir\n latest = \"\"\n for folder in folders:\n subfolder = folder.split(\"/\")[-2]\n if subfolder.startswith(\"prediction-\"):\n if subfolder > latest:\n latest = folder[:-1]\n return latest\n\n\nwhile True:\n predictions, state = get_batch_prediction_job(batch_job_id, True)\n if state != aip.JobState.JOB_STATE_SUCCEEDED:\n print(\"The job has not completed:\", state)\n if state == aip.JobState.JOB_STATE_FAILED:\n raise Exception(\"Batch Job Failed\")\n else:\n folder = get_latest_predictions(predictions)\n ! gsutil ls $folder/prediction*.csv\n\n ! gsutil cat $folder/prediction*.csv\n break\n time.sleep(60)", "_____no_output_____" ] ], [ [ "# Cleaning up\n\nTo clean up all GCP resources used in this project, you can [delete the GCP\nproject](https://cloud.google.com/resource-manager/docs/creating-managing-projects#shutting_down_projects) you used for the tutorial.\n\nOtherwise, you can delete the individual resources you created in this tutorial:\n\n- Dataset\n- Pipeline\n- Model\n- Endpoint\n- Batch Job\n- Custom Job\n- Hyperparameter Tuning Job\n- Cloud Storage Bucket", "_____no_output_____" ] ], [ [ "delete_dataset = True\ndelete_pipeline = True\ndelete_model = True\ndelete_endpoint = True\ndelete_batchjob = True\ndelete_customjob = True\ndelete_hptjob = True\ndelete_bucket = True\n\n# Delete the dataset using the Vertex fully qualified identifier for the dataset\ntry:\n if delete_dataset and \"dataset_id\" in globals():\n clients[\"dataset\"].delete_dataset(name=dataset_id)\nexcept Exception as e:\n print(e)\n\n# Delete the training pipeline using the Vertex fully qualified identifier for the pipeline\ntry:\n if delete_pipeline and \"pipeline_id\" in globals():\n clients[\"pipeline\"].delete_training_pipeline(name=pipeline_id)\nexcept Exception as e:\n print(e)\n\n# Delete the model using the Vertex fully qualified identifier for the model\ntry:\n if delete_model and \"model_to_deploy_id\" in globals():\n clients[\"model\"].delete_model(name=model_to_deploy_id)\nexcept Exception as e:\n print(e)\n\n# Delete the endpoint using the Vertex fully qualified identifier for the endpoint\ntry:\n if delete_endpoint and \"endpoint_id\" in globals():\n clients[\"endpoint\"].delete_endpoint(name=endpoint_id)\nexcept Exception as e:\n print(e)\n\n# Delete the batch job using the Vertex fully qualified identifier for the batch job\ntry:\n if delete_batchjob and \"batch_job_id\" in globals():\n clients[\"job\"].delete_batch_prediction_job(name=batch_job_id)\nexcept Exception as e:\n print(e)\n\n# Delete the custom job using the Vertex fully qualified identifier for the custom job\ntry:\n if delete_customjob and \"job_id\" in globals():\n clients[\"job\"].delete_custom_job(name=job_id)\nexcept Exception as e:\n print(e)\n\n# Delete the hyperparameter tuning job using the Vertex fully qualified identifier for the hyperparameter tuning job\ntry:\n if delete_hptjob and \"hpt_job_id\" in globals():\n clients[\"job\"].delete_hyperparameter_tuning_job(name=hpt_job_id)\nexcept Exception as e:\n print(e)\n\nif delete_bucket and \"BUCKET_NAME\" in globals():\n ! gsutil rm -r $BUCKET_NAME", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "BSD-2-Clause" ]

[ "BSD-2-Clause" ]

[ "BSD-2-Clause" ]

[ [ [ "# Project: Linear Regression\n\nReggie is a mad scientist who has been hired by the local fast food joint to build their newest ball pit in the play area. As such, he is working on researching the bounciness of different balls so as to optimize the pit. He is running an experiment to bounce different sizes of bouncy balls, and then fitting lines to the data points he records. He has heard of linear regression, but needs your help to implement a version of linear regression in Python.\n\n_Linear Regression_ is when you have a group of points on a graph, and you find a line that approximately resembles that group of points. A good Linear Regression algorithm minimizes the _error_, or the distance from each point to the line. A line with the least error is the line that fits the data the best. We call this a line of _best fit_.\n\nWe will use loops, lists, and arithmetic to create a function that will find a line of best fit when given a set of data.\n", "_____no_output_____" ], [ "## Part 1: Calculating Error", "_____no_output_____" ], [ "\nThe line we will end up with will have a formula that looks like:\n```\ny = m*x + b\n```\n`m` is the slope of the line and `b` is the intercept, where the line crosses the y-axis.\n\nCreate a function called `get_y()` that takes in `m`, `b`, and `x` and returns what the `y` value would be for that `x` on that line!\n", "_____no_output_____" ] ], [ [ "def get_y(m, b, x):\n y = m*x + b\n return y\n\nget_y(1, 0, 7) == 7\nget_y(5, 10, 3) == 25\n", "_____no_output_____" ] ], [ [ "\nReggie wants to try a bunch of different `m` values and `b` values and see which line produces the least error. To calculate error between a point and a line, he wants a function called `calculate_error()`, which will take in `m`, `b`, and an [x, y] point called `point` and return the distance between the line and the point.\n\nTo find the distance:\n1. Get the x-value from the point and store it in a variable called `x_point`\n2. Get the x-value from the point and store it in a variable called `y_point`\n3. Use `get_y()` to get the y-value that `x_point` would be on the line\n4. Find the difference between the y from `get_y` and `y_point`\n5. Return the absolute value of the distance (you can use the built-in function `abs()` to do this)\n\nThe distance represents the error between the line `y = m*x + b` and the `point` given.\n", "_____no_output_____" ] ], [ [ "def calculate_error(m, b, point):\n x_point, y_point = point\n y = m*x_point + b\n distance = abs(y - y_point)\n return distance\n", "_____no_output_____" ] ], [ [ "Let's test this function!", "_____no_output_____" ] ], [ [ "#this is a line that looks like y = x, so (3, 3) should lie on it. thus, error should be 0:\nprint(calculate_error(1, 0, (3, 3)))\n#the point (3, 4) should be 1 unit away from the line y = x:\nprint(calculate_error(1, 0, (3, 4)))\n#the point (3, 3) should be 1 unit away from the line y = x - 1:\nprint(calculate_error(1, -1, (3, 3)))\n#the point (3, 3) should be 5 units away from the line y = -x + 1:\nprint(calculate_error(-1, 1, (3, 3)))", "0\n1\n1\n5\n" ] ], [ [ "Great! Reggie's datasets will be sets of points. For example, he ran an experiment comparing the width of bouncy balls to how high they bounce:\n", "_____no_output_____" ] ], [ [ "datapoints = [(1, 2), (2, 0), (3, 4), (4, 4), (5, 3)]", "_____no_output_____" ] ], [ [ "The first datapoint, `(1, 2)`, means that his 1cm bouncy ball bounced 2 meters. The 4cm bouncy ball bounced 4 meters.\n\nAs we try to fit a line to this data, we will need a function called `calculate_all_error`, which takes `m` and `b` that describe a line, and `points`, a set of data like the example above.\n\n`calculate_all_error` should iterate through each `point` in `points` and calculate the error from that point to the line (using `calculate_error`). It should keep a running total of the error, and then return that total after the loop.\n", "_____no_output_____" ] ], [ [ "def calculate_all_error(m, b, points):\n total_error = 0\n for point in datapoints:\n point_error = calculate_error(m, b, point)\n total_error += point_error\n return total_error", "_____no_output_____" ] ], [ [ "Let's test this function!", "_____no_output_____" ] ], [ [ "#every point in this dataset lies upon y=x, so the total error should be zero:\ndatapoints = [(1, 1), (3, 3), (5, 5), (-1, -1)]\nprint(calculate_all_error(1, 0, datapoints))\n\n#every point in this dataset is 1 unit away from y = x + 1, so the total error should be 4:\ndatapoints = [(1, 1), (3, 3), (5, 5), (-1, -1)]\nprint(calculate_all_error(1, 1, datapoints))\n\n#every point in this dataset is 1 unit away from y = x - 1, so the total error should be 4:\ndatapoints = [(1, 1), (3, 3), (5, 5), (-1, -1)]\nprint(calculate_all_error(1, -1, datapoints))\n\n\n#the points in this dataset are 1, 5, 9, and 3 units away from y = -x + 1, respectively, so total error should be\n# 1 + 5 + 9 + 3 = 18\ndatapoints = [(1, 1), (3, 3), (5, 5), (-1, -1)]\nprint(calculate_all_error(-1, 1, datapoints))", "0\n4\n4\n18\n" ] ], [ [ "Great! It looks like we now have a function that can take in a line and Reggie's data and return how much error that line produces when we try to fit it to the data.\n\nOur next step is to find the `m` and `b` that minimizes this error, and thus fits the data best!\n", "_____no_output_____" ], [ "## Part 2: Try a bunch of slopes and intercepts!\n\nThe way Reggie wants to find a line of best fit is by trial and error. He wants to try a bunch of different slopes (`a` values) and a bunch of different intercepts (`b` values) and see which one produces the smallest error value for his dataset.\n\nUsing a list comprehension, let's create a list of possible `a` values to try. Make the list `possible_as` that goes from -10 to 10, in increments of 0.1.", "_____no_output_____" ], [ "The way Reggie wants to find a line of best fit is by trial and error. He wants to try a bunch of different slopes (`m` values) and a bunch of different intercepts (`b` values) and see which one produces the smallest error value for his dataset.\n\nUsing a list comprehension, let's create a list of possible `m` values to try. Make the list `possible_ms` that goes from -10 to 10, in increments of 0.1.\n\nHint (to view this hint, either double-click this cell or highlight the following white space): <font color=\"white\">you can go through the values in range(-100, 100) and multiply each one by 0.1</font>\n\n", "_____no_output_____" ] ], [ [ "possible_ms = [m * 0.1 for m in range(-100, 100)]", "_____no_output_____" ] ], [ [ "Now, let's make a list of `possible_bs` to check that would be the values from -20 to 20, in steps of 0.1:", "_____no_output_____" ] ], [ [ "possible_bs = [b * 0.1 for b in range(-200, 200)]", "_____no_output_____" ] ], [ [ "We are going to find the smallest error. First, we will make every possible `y = m*x + b` line by pairing all of the possible `m`s with all of the possible `b`s. Then, we will see which `y = m*x + b` line produces the smallest total error with the set of data stored in `datapoint`.\n\nFirst, create the variables that we will be optimizing:\n* `smallest_error` &mdash; this should start at infinity (`float(\"inf\")`) so that any error we get at first will be smaller than our value of `smallest_error`\n* `best_m` &mdash; we can start this at `0`\n* `best_b` &mdash; we can start this at `0`\n\nWe want to:\n* Iterate through each element `m` in `possible_ms`\n* For every `m` value, take every `b` value in `possible_bs`\n* If the value returned from `calculate_all_error` on this `m` value, this `b` value, and `datapoints` is less than our current `smallest_error`,\n* Set `best_m` and `best_b` to be these values, and set `smallest_error` to this error.\n\nBy the end of these nested loops, the `smallest_error` should hold the smallest error we have found, and `best_m` and `best_b` should be the values that produced that smallest error value.\n\nPrint out `best_m`, `best_b` and `smallest_error` after the loops.\n\n", "_____no_output_____" ] ], [ [ "datapoints = [(1, 2), (2, 0), (3, 4), (4, 4), (5, 3)]\nbest_error = float(\"inf\")\nbest_m = 0\nbest_b = 0\n\nfor m in possible_ms:\n for b in possible_bs:\n \t error = calculate_all_error(m, b, datapoints)\n \t if error < best_error:\n \t\t best_m = m\n \t\t best_b = b\n \t\t best_error = error\n \t \nprint(best_m, best_b, best_error)\n", "0.30000000000000004 1.7000000000000002 4.999999999999999\n" ] ], [ [ "## Part 3: What does our model predict?\n\nNow we have seen that for this set of observations on the bouncy balls, the line that fits the data best has an `m` of 0.3 and a `b` of 1.7:\n\n```\ny = 0.3x + 1.7\n```\n\nThis line produced a total error of 5.\n\nUsing this `m` and this `b`, what does your line predict the bounce height of a ball with a width of 6 to be?\nIn other words, what is the output of `get_y()` when we call it with:\n* m = 0.3\n* b = 1.7\n* x = 6", "_____no_output_____" ] ], [ [ "get_y(0.3, 1.7, 6)", "_____no_output_____" ] ], [ [ "Our model predicts that the 6cm ball will bounce 3.5m.\n\nNow, Reggie can use this model to predict the bounce of all kinds of sizes of balls he may choose to include in the ball pit!", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# Practical Examples of Interactive Visualizations in JupyterLab with Pixi.js and Jupyter Widgets\n\n# PyData Berlin 2018 - 2018-07-08\n\n# Jeremy Tuloup\n\n# [@jtpio](https://twitter.com/jtpio)\n# [github.com/jtpio](https://github.com/jtpio)\n# [jtp.io](https://jtp.io)", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# The Python Visualization Landscape (2017)\n\n\n\n\nSource:\n\n- [Jake VanderPlas: The Python Visualization Landscape PyCon 2017](https://www.youtube.com/watch?v=FytuB8nFHPQ)\n- [Source for the Visualization](https://github.com/rougier/python-visualization-landscape), by Nicolas P. Rougier", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# Motivation\n\n\n|Not This|This|\n|:--------------------------:|:-----------------------------------------:|\n| | |", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# JupyterLab - Pixi.js - Jupyter Widgets?", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# Prerequisites\n\n# * Jupyter Notebook\n# * Python", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# JupyterLab", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "\n\n## * Powerful 2D rendering engine written in JavaScript\n## * Abstraction on top of Canvas and WebGL\n\n# [Live Example!](http://localhost:4000)\n\n```javascript\nlet app = new PIXI.Application(800, 600, {backgroundColor : 0x1099bb});\ndocument.body.appendChild(app.view);\n\nlet bunny = PIXI.Sprite.fromImage('bunny.png')\n\nbunny.anchor.set(0.5);\n\nbunny.x = app.screen.width / 2;\nbunny.y = app.screen.height / 2;\n\napp.stage.addChild(bunny);\n\napp.ticker.add((delta) => {\n bunny.rotation += 0.1 * delta;\n});\n```", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# Jupyter Widgets\n\n\n\n[Open the image](./img/WidgetModelView.png)\n\n- Source: [https://ipywidgets.readthedocs.io/en/stable/examples/Widget%20Basics.html#Why-does-displaying-the-same-widget-twice-work?](https://ipywidgets.readthedocs.io/en/stable/examples/Widget%20Basics.html#Why-does-displaying-the-same-widget-twice-work?)", "_____no_output_____" ] ], [ [ "from ipywidgets import IntSlider\nslider = IntSlider(min=0, max=10)\nslider", "_____no_output_____" ], [ "slider", "_____no_output_____" ], [ "slider.value", "_____no_output_____" ], [ "slider.value = 2", "_____no_output_____" ] ], [ [ "# Tutorial to create your own\n\n## https://ipywidgets.readthedocs.io/en/stable/examples/Widget%20Custom.html\n\n# Libraries\n\n## bqplot\n\n\n\n## ipyleaflet\n\n\n\n## ipyvolume\n\n", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# Motivation: Very Custom Visualizations", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# Drawing Shapes on a Canvas", "_____no_output_____" ] ], [ [ "from ipyutils import SimpleShape", "_____no_output_____" ] ], [ [ "# Implementation\n\n## - [simple_shape.py](../ipyutils/simple_shape.py): defines the **SimpleShape** Python class\n## - [widget.ts](../src/simple_shapes/widget.ts): defines the **SimpleShapeModel** and **SimpleShapeView** Typescript classes", "_____no_output_____" ] ], [ [ "square = SimpleShape()\nsquare", "_____no_output_____" ], [ "square.rotate = True", "_____no_output_____" ] ], [ [ "# Level Up 🚀", "_____no_output_____" ] ], [ [ "from ipyutils import Shapes\nshapes = Shapes(n_shapes=100)\nshapes", "_____no_output_____" ], [ "shapes.shape", "_____no_output_____" ], [ "shapes.shape = 'square'", "_____no_output_____" ], [ "shapes.rotate = True", "_____no_output_____" ], [ "shapes.wobble = True", "_____no_output_____" ] ], [ [ "", "_____no_output_____" ], [ "# Visualizing Recursion with the Bermuda Triangle Puzzle\n\n", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# Motivation\n\n# * Solve the puzzle programmatically\n# * Verify a solution visually\n# * Animate the process", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# BermudaTriangle Widget", "_____no_output_____" ] ], [ [ "from ipyutils import TriangleAnimation, BermudaTriangle\ntriangles = TriangleAnimation()\ntriangles", "_____no_output_____" ] ], [ [ "", "_____no_output_____" ], [ "# What can we do with this widget?", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# Visualize Transitions\n\nFrom | To\n:--------------------------:|:-------------------------:\n | ", "_____no_output_____" ] ], [ [ "# states\nstate_0 = [None] * 16\nprint(state_0)", "[None, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None]\n" ], [ "state_1 = [[13, 1]] + [None] * 15\nprint(state_1)", "[[13, 1], None, None, None, None, None, None, None, None, None, None, None, None, None, None, None]\n" ], [ "state_2 = [[13, 1], [12, 0]] + [None] * 14\nprint(state_2)", "[[13, 1], [12, 0], None, None, None, None, None, None, None, None, None, None, None, None, None, None]\n" ] ], [ [ "# Example States and Animation", "_____no_output_____" ] ], [ [ "example_states = TriangleAnimation()\nbermuda = example_states.bermuda\nbermuda.states = [\n [None] * 16,\n [[7, 0]] + [None] * 15,\n [[7, 1]] + [None] * 15,\n [[7, 2]] + [None] * 15,\n [[7, 2], [0, 0]] + [None] * 14,\n [[7, 2], [0, 1]] + [None] * 14,\n [[i, 0] for i in range(16)],\n [[i, 1] for i in range(16)],\n]\nexample_states", "_____no_output_____" ] ], [ [ "", "_____no_output_____" ], [ "# Solver", "_____no_output_____" ] ], [ [ "from copy import deepcopy\n\nclass Solver(BermudaTriangle):\n\n def __init__(self, **kwargs):\n super().__init__(**kwargs)\n self.reset_state()\n \n def reset_state(self):\n self.board = [None] * self.N_TRIANGLES\n self.logs = [deepcopy(self.board)]\n self.it = 0\n \n def solve(self):\n '''\n Method to implement\n '''\n raise NotImplementedError()\n\n def log(self):\n self.logs.append(deepcopy(self.board))\n \n def found(self):\n return all(self.is_valid(i) for i in range(self.N_TRIANGLES))\n \n def save_state(self):\n self.permutation = self.board\n self.states = self.logs", "_____no_output_____" ] ], [ [ "# Valid Permutation - is_valid()", "_____no_output_____" ] ], [ [ "help(Solver.is_valid)", "Help on function is_valid in module ipyutils.bermuda:\n\nis_valid(self, i)\n Parameters\n ----------\n \n i: int\n Position of the triangle to check, between 0 and 15 (inclusive)\n \n Returns\n -------\n valid: bool\n True if the triangle at position i doesn't have any conflict\n False otherwise\n\n" ] ], [ [ "```python\nsolver.is_valid(7)\n# False\n```\n", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# First Try: Random Permutations", "_____no_output_____" ] ], [ [ "import random\n\nclass RandomSearch(Solver):\n \n def solve(self):\n random.seed(42)\n self.reset_state()\n for i in range(200):\n self.board = random.sample(self.permutation, self.N_TRIANGLES)\n self.log()\n if self.found():\n print('Found!')\n return True\n return False", "_____no_output_____" ], [ "%%time\n\nsolver = RandomSearch()\nres = solver.solve()\nsolver.save_state()", "CPU times: user 93.8 ms, sys: 3.94 ms, total: 97.7 ms\nWall time: 97.5 ms\n" ], [ "rnd = TriangleAnimation()\nrnd.bermuda.title = 'Random Search'\nrnd.bermuda.states = solver.states\nrnd", "_____no_output_____" ] ], [ [ "", "_____no_output_____" ], [ "# Better: Brute Force using Recursion", "_____no_output_____" ] ], [ [ "class RecursiveSolver(Solver):\n \n def solve(self):\n self.used = [False] * self.N_TRIANGLES\n self.reset_state()\n self._place(0)\n return self.board\n \n def _place(self, i):\n self.it += 1\n if i == self.N_TRIANGLES:\n return True\n \n for j in range(self.N_TRIANGLES - 1, -1, -1):\n if self.used[j]:\n # piece number j already used\n continue\n \n self.used[j] = True\n \n for rot in range(3):\n # place the piece on the board\n self.board[i] = (j, rot)\n self.log()\n\n # stop the recursion if the current configuration\n # is not valid or a solution has been found\n if self.is_valid(i) and self._place(i + 1):\n return True\n\n # remove the piece from the board\n self.board[i] = None\n self.used[j] = False\n self.log()\n \n return False", "_____no_output_____" ], [ "%%time\n\nsolver = RecursiveSolver()\nres = solver.solve()\nif solver.found():\n print('Solution found!')\n print(f'{len(solver.logs)} steps')\n solver.save_state()\nelse:\n print('No solution found')", "Solution found!\n10753 steps\nCPU times: user 2.3 s, sys: 10.1 ms, total: 2.31 s\nWall time: 2.32 s\n" ], [ "recursion = TriangleAnimation()\nrecursion.bermuda.title = 'Recursive Search'\nrecursion.bermuda.states = solver.states\nrecursion", "_____no_output_____" ] ], [ [ "", "_____no_output_____" ], [ "# More details for this example\n\n## * In depth walkthrough on how to create a Jupyter Widget in the notebook\n## * [p5.js in the Jupyter Notebook for custom interactive visualizations](https://github.com/jtpio/p5-jupyter-notebook/blob/master/puzzle.ipynb)\n## * Using p5.js instead of Pixi.js, but similar concepts\n## * Source: [github.com/jtpio/p5-jupyter-notebook](https://github.com/jtpio/p5-jupyter-notebook)\n## * [Run on Binder](https://mybinder.org/v2/gh/jtpio/p5-jupyter-notebook/master?filepath=puzzle.ipynb)", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# Recap\n\n## * Custom interactive animations with Pixi.js\n## * Leverage the JavaScript ecosystem in JupyterLab", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# Applications\n\n## * Visual debugging and understanding\n## * Teaching and education, learning by doing\n## * Combine JavaScript games with data\n\n# Downside\n\n## * Requires some effort to build the visualizations in TypeScript / JavaScript", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# References\n\n## Presentations\n\n### - [Jake VanderPlas: The Python Visualization Landscape PyCon 2017](https://www.youtube.com/watch?v=FytuB8nFHPQ)\n### - [PyData London 2016: Sylvain Corlay - Interactive Visualization in Jupyter with Bqplot and Interactive Widgets](https://www.youtube.com/watch?v=eVET9IYgbao)\n### - [PLOTCON 2017: Sylvain Corlay, Interactive Data Visualization in JupyterLab with Jupyter](https://www.youtube.com/watch?v=p7Hr54VhOp0)\n### - [PyData Amsterdam 2017: Maarten Breddels | A billion stars in the Jupyter Notebook](https://www.youtube.com/watch?v=bP-JBbjwLM8)\n\n## Widgets\n\n### - [Building a Custom Widget Tutorial](https://ipywidgets.readthedocs.io/en/stable/examples/Widget%20Custom.html)\n### - [Authoring Custom Jupyter Widgets](https://blog.jupyter.org/authoring-custom-jupyter-widgets-2884a462e724)\n### - [p5.js in the Jupyter Notebook for custom interactive visualizations](https://github.com/jtpio/p5-jupyter-notebook/blob/master/puzzle.ipynb)\n### - [pythreejs](https://github.com/jovyan/pythreejs): Implemented as an Jupyter Widget\n### - [bqplot](https://github.com/bloomberg/bqplot): Great library for interactive data exploration\n### - [ipyvolume](https://github.com/maartenbreddels/ipyvolume): 3d plotting for Python in the Jupyter Notebook\n### - [ipyleaflet](https://github.com/jupyter-widgets/ipyleaflet): interactive maps in the Jupyter notebook", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "# Questions?\n\n## [@jtpio](https://twitter.com/jtpio)\n\n## [github.com/jtpio](https://github.com/jtpio)\n\n## [jtp.io](jtp.io)", "_____no_output_____" ], [ "", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Regressão linear\n\n\n## **TOC:**\n\nNa aula de hoje, vamos explorar os seguintes tópicos em Python:\n\n- 1) [Introdução](#intro)\n- 2) [Regressão linear simples](#reglinear)\n- 3) [Regressão linear múltipla](#multireglinear)\n- 4) [Tradeoff viés-variância](#tradeoff)", "_____no_output_____" ] ], [ [ "# importe as principais bibliotecas de análise de dados\nimport pandas as pd\nimport numpy as np\n\nimport matplotlib.pyplot as plt\nimport seaborn as sns", "_____no_output_____" ] ], [ [ "____\n____\n____", "_____no_output_____" ], [ "## 1) **Introdução** <a class=\"anchor\" id=\"intro\"></a>\n\nImagine que você que vender sua casa.\n\nVocê sabe os atributos da sua casa: quantos cômodos têm, quantos carros cabem na garagem, qual é a área construída, qual sua localidade, etc.\n\nAgora, a pergunta é: qual seria o melhor preço pra você colocá-la a venda, ou seja, quanto de fato ela vale?\n\nVocê pode solicitar a avaliação de um corretor de imóveis (contanto com a experiência dele), ou então...\n\n...fazer um modelo de **Machine Learning**, que, com base nos atributos e preços de diversas outras casas, pode fazer uma **predição** sobre o preço adequado da sua casa!\n\nPara resolver este problema, podemos utilizar um dos mais simples e importantes algoritmos de machine learning: a Regressão Linear!\n\n____", "_____no_output_____" ], [ "Para introduzirmos as ideias, vamos usar um [dataset de preço de casas](https://www.kaggle.com/c/house-prices-advanced-regression-techniques/data).\n\nEsta base de dados contém **70 features** (+ 1 ID), que são as características de cada uma das casas listadas; e **1 target**, que é o preço pelo qual aquela casa foi vendida.\n\nPara o significado de cada uma das features, e os valores que elas podem assumir, veja a página acima.\n\n**Vamos ler a base e começar a explorá-la!**", "_____no_output_____" ] ], [ [ "df = pd.read_csv(\"data/house_prices/house_price.csv\")", "_____no_output_____" ] ], [ [ "Por enquanto, não vamos nos preocupar com os dados missing, pois vamos usar apenas uma feature no nosso modelo inicial.\n\nAproveite para depois explorar os dados da forma que quiser!\n\nPor enquanto, vamos dar uma olhada na coluna target!", "_____no_output_____" ], [ "Fica evidente que a distribuição é desviada para a direita.\n\nVamos tentar alterar isso na próximas versões do modelo para ver se teremos ganhos de performance!\n\nPor enquanto, seguimos assim.", "_____no_output_____" ], [ "Parece que a variável de área construída ```GrLivArea```) é uma forte candidata a **explicar** o preço das casas, pois vemos calaramente uma correlação entre as variáveis!\n\nMas note que há claramente dois outliers... ", "_____no_output_____" ], [ "Vamos agora iniciar a construção de um modelo bem simples, que utilize a variável GrLivArea para predizer o preço!", "_____no_output_____" ], [ "___\n___\n___", "_____no_output_____" ], [ "## 2) **Regressão linear simples** <a class=\"anchor\" id=\"reglinear\"></a>\n\nApesar de alguns outliers, parece bem adequado que os pontos plotados acima sejam descritos por uma reta, não é mesmo?\n\nOu, melhor dizendo: **a variável GrLivArea parece estar relacionada ao target SalePrice linearmente!**\n\nPara modelarmos esta relação, vamos conhecer o modelo de **Regressão Linear Simples**.\n\nComo o próprio nome diz, o modelo de Regressão Linear será **uma reta (polinômio linear)**, que melhor se ajusta aos seus dados!\n\nO modelo de **Regressão Linear Simples** será uma linha reta que relaciona Y (o preço da casa) e X (os atributos da casa). \n\nSe utilizarmos **apenas um atributo** (como, por exemplo, a área construída), temos uma **Regressão Linear Simples**, e nosso modelo é:\n\n$$ y = b_0 + b_1 X $$\n\nNeste caso, o modelo tem dois coeficientes a serem determinados: $b_0$ (intercepto ou coeficiente linear) e $b_1$ (coeficiente angular). \n\nO algoritmo do estimador é utilizado justamente para encontrarmos os coeficientes $b_0$ e $b_1$ **que melhor se ajustam aos dados!**\n\nPara fazer isso, pode-se utilizar o método dos **mínimos quadrados** ou então o **gradiente descendente**.\n\nMas não nos preocuparemos com os detalhes do treinamento: usaremos o sklearn para isso!\n\nVamos começar?", "_____no_output_____" ], [ "Agora que o modelo está treinado, podemos dar uma olhada nos coeficientes que foram encontrados!", "_____no_output_____" ], [ "Como interpretamos este resultado?\n\nO nosso modelo final é dado por:\n\n$$ y = 1562.01 + 118.61 \\times \\text{GrLiveArea}$$\n\nIsto quer dizer que:\n\n> Aumentando a variável \"GrLiveArea\" em uma unidade faz com que o preço seja aumentado em USD 118.6!\n\n> O preço mínimo a ser pago, independente da área construída, é de 1562.01!", "_____no_output_____" ], [ "Podemos visualizar o modelo treinado, neste caso:", "_____no_output_____" ], [ "Fazendo uma previsão:", "_____no_output_____" ], [ "Ou ainda:", "_____no_output_____" ], [ "É raro que consigamos visualizar nosso modelo final como fizemos acima, mas no caso da regressão linear simples, temos essa sorte! :)\n\nVamos agora fazer algumas previsões!", "_____no_output_____" ], [ "Agora que temos o modelo treinado e algumas previsões, como avaliamos a performance do modelo?\n\nPara isso, podemos dar uma olhada nos **resíduos** das predições! Os resíduos nada mais são do que **os erros do modelo**, ou seja, **a diferença entre cada valor predito e o valor real**, para **os dados de teste!**. Isto é,\n\n$$R(y_i) = y_i - \\hat{y}_i $$", "_____no_output_____" ], [ "O caso 100% ideal seria $y_i = \\hat{y}_i$, o que produziria uma reta exata!\n\nQuanto mais \"espalhados\" estiverem os pontos em torno da reta, em geral **pior é o modelo**, pois ele está errando mais!\n\nUma forma de quantificar isso através de uma métrica conhecida como **$R^2$**, o **coeficiente de determinação**.\n\nEste coeficiente indica **o quão próximos os dados estão da reta ajustada**. Por outro lado, o $R^2$ representa a porcentagem de variação na resposta que é explicada pelo modelo.\n\n$$R^2 = 1 - \\frac{\\sum_{i=1}^n(y_i-\\hat{y}_i)^2}{\\sum_{i=1}^n(y_i-\\bar{y})^2}$$\n\nÉ possível utilizar o $R^2$ nos dados de treino, mas isso não é tão significante, devido ao overfitting, que discutiremos a seguir. Mais sgnificativo é calcularmos o $R^2$ nos dados de teste como faremos a seguir. Essa métrica equivale, portanto, **ao gráfico que fizemos acima!**\n", "_____no_output_____" ], [ "Outra coisa importante é que os resíduos sejam **normalmente distribuídos**.\n\nSe esse não for o caso, é muito importante que você reveja se o modelo escolhido de fato é adequado ao seu problema!", "_____no_output_____" ], [ "Além dos resíduos, existem três principais **métricas de avaliação** do modelo de regressão linear:", "_____no_output_____" ], [ "**Mean Absolute Error** (MAE) é a média do valor absoluto de todos os resíduos (erros):\n\n$$\\frac 1n\\sum_{i=1}^n|y_i-\\hat{y}_i|$$\n\n**Mean Squared Error** (MSE) é a média dos erros quadrados:\n\n$$\\frac 1n\\sum_{i=1}^n(y_i-\\hat{y}_i)^2$$\n\n**Root Mean Squared Error** (RMSE) é a raiz quadrada da média dos erros quadrados:\n\n$$\\sqrt{\\frac 1n\\sum_{i=1}^n(y_i-\\hat{y}_i)^2}$$\n\nComparando as métricas:\n\n- **MAE** é a mais simples de entender, mas ela penaliza mais erros menores;\n- **MSE** é a métrica mais popular, pois essa métrica penaliza mais erros maiores, o que faz mais sentido em aplicações reais.\n- **RMSE** é ainda mais popular, pois esta métrica está nas mesmas unidades que o target.\n\nEstas métricas todas podem ser utilizadas como **funções de custo** a serem minimizadas pelo algoritmo do estimador.", "_____no_output_____" ], [ "___", "_____no_output_____" ], [ "## 3) **Regressão linear múltipla** <a class=\"anchor\" id=\"multireglinear\"></a>\n\n\nO modelo que fizemos acima considera uma única feature como preditora do preço da casa.\n\nMas temos outras 78 dessas features! Será que não há mais informação útil em todas essas outras variáveis?\n\nEm geral, sim! É natural que esperemos que **mais variáveis** tragam **mais informações** ao modelo, e, portanto, o torne mais preciso!\n\nPara incorporar estas outras variáveis ao modelo, é muito simples! \n\nPodemos passar a utilizar outros atributos (como o número de cômodos, qual é a renda média da vizinhança, etc.), e neste caso teremos uma **Regressão Linear Múltipla**, que nada mais é que a seguinte equação:\n\n$$ y = b_0 + b_1 X_1 + b_2 X_2 + \\cdots + b_n X_n $$\n\nNeste caso, além de $b_0$ e $b_1$, temos também outros coeficientes, um pra cada uma das $n$ features que escolhermos!\n\nModelos de regressão múltipla são potencialmente mais precisos, mas há também um lado ruim: nós perdemos a **possibilidade de visualização**. Agora, não temos mais uma reta, mas sim um **hiperplano** que relaciona todas as features com o target!\n\n<center><img src=\"https://miro.medium.com/max/1120/0*rGSfRsMjiQeG5jof.png\" width=500></center>\n\nVamos construir esse modelo?", "_____no_output_____" ], [ "Observação: a coluna \"Id\" traz apenas um número de identificação arbitrário que não deve ser correlacionado com o target. Portanto, vamos desconsiderar esta coluna de nosso modelo!", "_____no_output_____" ], [ "A performance do modelo melhorou?\n\nSerá que dá pra melhorar mais?\n\nOpções:\n\n- tentar apenas um subconjunto de features: **feature selection**\n\n\n- passar a utilizar as features categóricas: **feature engeneering**", "_____no_output_____" ], [ "---", "_____no_output_____" ], [ "## 4) **Tradeoff viés-variância** <a class=\"anchor\" id=\"tradeoff\"></a>\n\nVeremos agora um dos conceitos mais importantes em apredizado de maquina.\n\nMuitas vezes alguns modelos têm 100% de acerto nos dados de **treino**, mas **na base de teste** a performance cai para menos de 50%.\n\nIsso pode acontecer porque o modelo fica **especialista apenas no conjunto de treino**, não conseguindo **generalizar os padrões para além dos dados vistos**.\n\n<center><img src=\"https://miro.medium.com/max/1125/1*_7OPgojau8hkiPUiHoGK_w.png\" width=800></center>\n\nO overfitting está intimamente ligado com o conceito de **viés** (bias) e **variância** (variance):\n\n>**Viés**<br>\nÉ a diferença entre o que o modelo prediz, e o valor correto a ser predito.<br>\nModelos com alto viés são muito simples, de modo a **não conseguir capturar as relações que os dados de treino exibem** (underfit).<br>\nIssso faz com que ambos os erros de treino e de teste sejam altos.\n<br><br>\nEm outras palavras:<br>\n**Incapacidade de um modelo de capturar a verdadeira relação entre features e target**\n\n\n> **Variância**<br>\nVariância se refere à variabilidade das predições de um modelo.<br>\nModelos com alta variância são muito complexos, por **aprenderem demais as relações exibidas nos dados de treino** (overfit).<br>\nIsso faz com que os erros de treino sejam baixos, mas os erros de teste sejam altos.\n<br><br>\nEm outras palavras:<br>\n**Incapacidade de um modelo performar bem em outros datasets diferentes do usado no treinamento**. \n\n<center><img src=\"https://www.learnopencv.com/wp-content/uploads/2017/02/Bias-Variance-Tradeoff-In-Machine-Learning-1.png\" width=500></center>\n\n<center><img src=\"https://miro.medium.com/max/1494/1*C7ZKM93QVdpeSCGbF5TjIg.png\" width=800></center>\n\nPara demonstrar overfit ser usado o conjuto de teste [anscombe](https://en.wikipedia.org/wiki/Anscombe%27s_quartet)", "_____no_output_____" ] ], [ [ "df_anscombe = sns.load_dataset('anscombe')\n\ndf_anscombe.groupby(\"dataset\").agg({\"mean\", \"std\"})", "_____no_output_____" ] ], [ [ "Vamos supor que este dado represente valores de medições de um sensor, porém o sensor teve um pequeno problema durante a medição.\n\nPodemos perceber facilmente qual é este erro, e qual seria a função de regreesão para este sensor com os dados validos: **regressão linear**.", "_____no_output_____" ], [ "Perceba que a função linear encontrar já aprensenta um padrão muito similiar aos dados, porém um ponto error faz com que ela não tenha um resultado otimo.\n\nPodemos utilizar regressões polinomiais, que possuem ordem maiores que 1, para tentar diminuir o erro da regressão, obtendo uma equação do formato.\n\n$$\\hat{y}_{i} = \\beta_{1} + \\beta_{2} x_{i} + \\beta_{3} {x_{i}}^{2} + \\cdots + \\beta_{6} {x_{i}}^{6}$$\n\nPara criar modelos polinomiaus com o sklearn, [dê uma olhada aqui](https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.PolynomialFeatures.html)", "_____no_output_____" ], [ "Ao utilizarmos uma regressão de ordem 6 percebemos que ela se ajusta ao valor com erro, porém ela **se distancia da regressão que realmente representa os dados**. \n\nTentar **aprender o erro faz com ela com ela não aprenda a função real**. \n\nIsto acontece pois ela se **super ajustou aos dados de treino, se distanciando dos dados reais**. ", "_____no_output_____" ], [ "__Como garantir que nosso modelo não está sofrendo de overfitting?__\n\nNaturalmente, essa é uma pergunta de extrema importância, especialmente no contexto de **Redes neurais**. [Veja aqui](https://towardsdatascience.com/8-simple-techniques-to-prevent-overfitting-4d443da2ef7d) e [aqui](https://towardsdatascience.com/dont-overfit-how-to-prevent-overfitting-in-your-deep-learning-models-63274e552323) algumas discussões.\n\nNa prática: **jamais se apegue à peformance de treino!!**. O que queremos otimizar sempre será a performance **avaliada nos dados de teste**. Assim, garantimos que uma boa performance não é produto do overfitting!", "_____no_output_____" ], [ "--- ", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "from oda_api.api import DispatcherAPI\nfrom oda_api.plot_tools import OdaImage,OdaLightCurve\nfrom oda_api.data_products import BinaryData\nimport os\nfrom astropy.io import fits\nimport numpy as np\nfrom numpy import sqrt\nimport matplotlib.pyplot as plt\n%matplotlib inline", "_____no_output_____" ], [ "source_name='3C 279'\nra=194.046527\ndec=-5.789314\nradius=10.\nTstart='2003-03-15T00:00:00'\nTstop='2018-03-15T00:00:00'\nE1_keV=30.\nE2_keV=100.\nhost='www.astro.unige.ch/cdci/astrooda/dispatch-data'\nrebin=10 # minimal significance in energy bin, for spectral plotting", "_____no_output_____" ], [ "try: input = raw_input\nexcept NameError: pass\ntoken=input() # token for restricted access server\ncookies=dict(_oauth2_proxy=token)\ndisp=DispatcherAPI(host=host)", "\n" ], [ "disp=DispatcherAPI(host=host)", "_____no_output_____" ], [ "import requests\nurl=\"https://www.astro.unige.ch/cdci/astrooda/dispatch-data/gw/timesystem/api/v1.0/scwlist/cons/\"\ndef queryxtime(**args):\n params=Tstart+'/'+Tstop+'?&ra='+str(ra)+'&dec='+str(dec)+'&radius='+str(radius)+'&min_good_isgri=1000'\n print(url+params)\n return requests.get(url+params,cookies=cookies).json()", "_____no_output_____" ], [ "scwlist=queryxtime()\nm=len(scwlist)\npointings_osa10=[]\npointings_osa11=[]\nfor i in range(m):\n if scwlist[i][-2:]=='10':\n if(int(scwlist[i][:4])<1626):\n pointings_osa10.append(scwlist[i]+'.001')\n else:\n pointings_osa11.append(scwlist[i]+'.001')\n#else:\n# pointings=np.genfromtxt('scws_3C279_isgri_10deg.txt', dtype='str')\nm_osa10=len(pointings_osa10)\nm_osa11=len(pointings_osa11)", "https://www.astro.unige.ch/cdci/astrooda/dispatch-data/gw/timesystem/api/v1.0/scwlist/cons/2003-03-15T00:00:00/2018-03-15T00:00:00?&ra=194.046527&dec=-5.789314&radius=10.0&min_good_isgri=1000\n" ], [ "scw_lists_osa10=[]\nscw_lists_osa11=[]\ncount=0\nscw_string=''\nfor i in range(m_osa10):\n if count<50:\n scw_string=scw_string+str(pointings_osa10[i])+','\n count+=1\n else:\n scw_lists_osa10.append(scw_string[:-1])\n count=0\n scw_string=str(pointings_osa10[i])+','\nscw_lists_osa10.append(scw_string[:-1])\nprint(len(scw_lists_osa10))\ncount=0\nscw_string=''\nfor i in range(m_osa11):\n if count<50:\n scw_string=scw_string+str(pointings_osa11[i])+','\n count+=1\n else:\n scw_lists_osa11.append(scw_string[:-1])\n count=0\n scw_string=str(pointings_osa11[i])+','\nscw_lists_osa11.append(scw_string[:-1])\nprint(len(scw_lists_osa11))", "22\n2\n" ], [ "data=disp.get_product(instrument='isgri',\n product='isgri_image',\n scw_list=scw_lists_osa10[0],\n E1_keV=E1_keV,\n E2_keV=E2_keV,\n osa_version='OSA10.2',\n RA=ra,\n DEC=dec,\n detection_threshold=3.5,\n product_type='Real')", "007700160010.001,007700170010.001,007700180010.001,007700190010.001,007700200010.001,007700210010.001,007700220010.001,007700230010.001,007700240010.001,007700250010.001,007700260010.001,007700270010.001,007700280010.001,007700290010.001,007700300010.001,007700310010.001,007700320010.001,007700330010.001,007700340010.001,007700350010.001,007700360010.001,007700370010.001,007700380010.001,007700390010.001,007700400010.001,007700410010.001,007700460010.001,007700470010.001,007700480010.001,007700490010.001,007700500010.001,007700510010.001,007700520010.001,007700530010.001,007700540010.001,007700550010.001,007700560010.001,007700570010.001,007700580010.001,007700590010.001,007700600010.001,007700660010.001,007700670010.001,007800020010.001,007800030010.001,007800040010.001,007800050010.001,007800060010.001,007800070010.001,007800080010.001\n- waiting for remote response, please wait run_analysis https://www.astro.unige.ch/cdci/astrooda/dispatch-data\nscw_list 007700160010.001,007700170010.001,007700180010.001,007700190010.001,007700200010.001,007700210010.001,007700220010.001,007700230010.001,007700240010.001,007700250010.001,007700260010.001,007700270010.001,007700280010.001,007700290010.001,007700300010.001,007700310010.001,007700320010.001,007700330010.001,007700340010.001,007700350010.001,007700360010.001,007700370010.001,007700380010.001,007700390010.001,007700400010.001,007700410010.001,007700460010.001,007700470010.001,007700480010.001,007700490010.001,007700500010.001,007700510010.001,007700520010.001,007700530010.001,007700540010.001,007700550010.001,007700560010.001,007700570010.001,007700580010.001,007700590010.001,007700600010.001,007700660010.001,007700670010.001,007800020010.001,007800030010.001,007800040010.001,007800050010.001,007800060010.001,007800070010.001,007800080010.001\nE1_keV 30.0\nE2_keV 100.0\nosa_version OSA10.2\nRA 194.046527\nDEC -5.789314\ndetection_threshold 3.5\ninstrument isgri\nproduct_type isgri_image\nquery_type Real\noff_line (False,)\nquery_status ('new',)\nverbose (False,)\nsession_id 9BBCH4B0GX5I0ZJJ\ndry_run (False,)\napi True\n\n\nquery done succesfully!\n" ], [ "data.dispatcher_catalog_1.table", "_____no_output_____" ], [ "FLAG=0\ntorm=[]\nfor ID,n in enumerate(data.dispatcher_catalog_1.table['src_names']):\n if(n[0:3]=='NEW'):\n torm.append(ID)\n if(n==source_name):\n FLAG=1\ndata.dispatcher_catalog_1.table.remove_rows(torm)\nnrows=len(data.dispatcher_catalog_1.table['src_names'])", "_____no_output_____" ], [ "if FLAG==0:\n data.dispatcher_catalog_1.table.add_row((0,'3C 279',0,ra,dec,0,2,0,0))\n", "_____no_output_____" ], [ "api_cat=data.dispatcher_catalog_1.get_api_dictionary()", "_____no_output_____" ], [ "spectrum_results=[]\nfor i in range(len(scw_lists_osa10)):\n print(i)\n data=disp.get_product(instrument='isgri',\n product='isgri_spectrum',\n scw_list=scw_lists_osa10[i],\n query_type='Real',\n osa_version='OSA10.2',\n RA=ra,\n DEC=dec,\n product_type='Real',\n selected_catalog=api_cat)\n spectrum_results.append(data)", "0\n007700160010.001,007700170010.001,007700180010.001,007700190010.001,007700200010.001,007700210010.001,007700220010.001,007700230010.001,007700240010.001,007700250010.001,007700260010.001,007700270010.001,007700280010.001,007700290010.001,007700300010.001,007700310010.001,007700320010.001,007700330010.001,007700340010.001,007700350010.001,007700360010.001,007700370010.001,007700380010.001,007700390010.001,007700400010.001,007700410010.001,007700460010.001,007700470010.001,007700480010.001,007700490010.001,007700500010.001,007700510010.001,007700520010.001,007700530010.001,007700540010.001,007700550010.001,007700560010.001,007700570010.001,007700580010.001,007700590010.001,007700600010.001,007700660010.001,007700670010.001,007800020010.001,007800030010.001,007800040010.001,007800050010.001,007800060010.001,007800070010.001,007800080010.001\n- waiting for remote response, please wait run_analysis https://www.astro.unige.ch/cdci/astrooda/dispatch-data\nscw_list 007700160010.001,007700170010.001,007700180010.001,007700190010.001,007700200010.001,007700210010.001,007700220010.001,007700230010.001,007700240010.001,007700250010.001,007700260010.001,007700270010.001,007700280010.001,007700290010.001,007700300010.001,007700310010.001,007700320010.001,007700330010.001,007700340010.001,007700350010.001,007700360010.001,007700370010.001,007700380010.001,007700390010.001,007700400010.001,007700410010.001,007700460010.001,007700470010.001,007700480010.001,007700490010.001,007700500010.001,007700510010.001,007700520010.001,007700530010.001,007700540010.001,007700550010.001,007700560010.001,007700570010.001,007700580010.001,007700590010.001,007700600010.001,007700660010.001,007700670010.001,007800020010.001,007800030010.001,007800040010.001,007800050010.001,007800060010.001,007800070010.001,007800080010.001\nquery_type Real\nosa_version OSA10.2\nRA 194.046527\nDEC -5.789314\nselected_catalog {\"cat_frame\": \"fk5\", \"cat_coord_units\": \"deg\", \"cat_column_list\": [[1, 30, 50, 112, 0], [\"3C 273\", \"IGR J12391-1612\", \"NGC 4593\", \"SWIFT J1238.6+0928\", \"3C 279\"], [12.098819732666016, 4.082015037536621, 13.66150951385498, 4.295688629150391, 0.0], [187.2660675048828, 189.77621459960938, 189.91522216796875, 189.67724609375, 194.046527], [2.045917510986328, -16.179750442504883, -5.358253479003906, 9.474955558776855, -5.789314], [-32768, -32768, -32768, -32768, 0], [2, 2, 2, 1, 2], [0, 0, 0, 0, 0], [0.0002800000074785203, 0.00016999999934341758, 0.0002800000074785203, 0.07566666603088379, 0.0]], \"cat_column_names\": [\"meta_ID\", \"src_names\", \"significance\", \"ra\", \"dec\", \"NEW_SOURCE\", \"ISGRI_FLAG\", \"FLAG\", \"ERR_RAD\"], \"cat_column_descr\": [[\"meta_ID\", \"<i8\"], [\"src_names\", \"<U18\"], [\"significance\", \"<f8\"], [\"ra\", \"<f8\"], [\"dec\", \"<f8\"], [\"NEW_SOURCE\", \"<i8\"], [\"ISGRI_FLAG\", \"<i8\"], [\"FLAG\", \"<i8\"], [\"ERR_RAD\", \"<f8\"]], \"cat_lat_name\": \"dec\", \"cat_lon_name\": \"ra\"}\ninstrument isgri\nproduct_type isgri_spectrum\noff_line (False,)\nquery_status ('new',)\nverbose (False,)\nsession_id Y4FZHGU8XTMLNQ5J\ndry_run (False,)\napi True\nthe job has been submitted on the remote server\n / the job is working remotely, please wait status=done - job_id=-283616046116470052 0052 \n\nquery done succesfully!\n1\n007800090010.001,007800100010.001,007800110010.001,007800120010.001,007800130010.001,007800140010.001,007800150010.001,007800160010.001,007800170010.001,007800180010.001,007800190010.001,007800200010.001,007800210010.001,007800220010.001,007800230010.001,007800240010.001,007800250010.001,007800260010.001,007800270010.001,007800280010.001,007800290010.001,007800300010.001,007800310010.001,007800320010.001,007800330010.001,007800340010.001,007800350010.001,007800360010.001,007800370010.001,007800380010.001,007800390010.001,007800400010.001,007800410010.001,007800420010.001,007800430010.001,007800440010.001,007800450010.001,007800460010.001,008900050010.001,008900060010.001,008900070010.001,008900080010.001,008900110010.001,008900120010.001,008900130010.001,008900140010.001,008900150010.001,008900160010.001,008900170010.001,008900230010.001,008900240010.001\n- waiting for remote response, please wait run_analysis https://www.astro.unige.ch/cdci/astrooda/dispatch-data\nscw_list 007800090010.001,007800100010.001,007800110010.001,007800120010.001,007800130010.001,007800140010.001,007800150010.001,007800160010.001,007800170010.001,007800180010.001,007800190010.001,007800200010.001,007800210010.001,007800220010.001,007800230010.001,007800240010.001,007800250010.001,007800260010.001,007800270010.001,007800280010.001,007800290010.001,007800300010.001,007800310010.001,007800320010.001,007800330010.001,007800340010.001,007800350010.001,007800360010.001,007800370010.001,007800380010.001,007800390010.001,007800400010.001,007800410010.001,007800420010.001,007800430010.001,007800440010.001,007800450010.001,007800460010.001,008900050010.001,008900060010.001,008900070010.001,008900080010.001,008900110010.001,008900120010.001,008900130010.001,008900140010.001,008900150010.001,008900160010.001,008900170010.001,008900230010.001,008900240010.001\nquery_type Real\nosa_version OSA10.2\nRA 194.046527\nDEC -5.789314\nselected_catalog {\"cat_frame\": \"fk5\", \"cat_coord_units\": \"deg\", \"cat_column_list\": [[1, 30, 50, 112, 0], [\"3C 273\", \"IGR J12391-1612\", \"NGC 4593\", \"SWIFT J1238.6+0928\", \"3C 279\"], [12.098819732666016, 4.082015037536621, 13.66150951385498, 4.295688629150391, 0.0], [187.2660675048828, 189.77621459960938, 189.91522216796875, 189.67724609375, 194.046527], [2.045917510986328, -16.179750442504883, -5.358253479003906, 9.474955558776855, -5.789314], [-32768, -32768, -32768, -32768, 0], [2, 2, 2, 1, 2], [0, 0, 0, 0, 0], [0.0002800000074785203, 0.00016999999934341758, 0.0002800000074785203, 0.07566666603088379, 0.0]], \"cat_column_names\": [\"meta_ID\", \"src_names\", \"significance\", \"ra\", \"dec\", \"NEW_SOURCE\", \"ISGRI_FLAG\", \"FLAG\", \"ERR_RAD\"], \"cat_column_descr\": [[\"meta_ID\", \"<i8\"], [\"src_names\", \"<U18\"], [\"significance\", \"<f8\"], [\"ra\", \"<f8\"], [\"dec\", \"<f8\"], [\"NEW_SOURCE\", \"<i8\"], [\"ISGRI_FLAG\", \"<i8\"], [\"FLAG\", \"<i8\"], [\"ERR_RAD\", \"<f8\"]], \"cat_lat_name\": \"dec\", \"cat_lon_name\": \"ra\"}\ninstrument isgri\nproduct_type isgri_spectrum\noff_line (False,)\nquery_status ('new',)\nverbose (False,)\nsession_id CQYQM9B2TEQLIQP0\ndry_run (False,)\napi True\nthe job has been submitted on the remote server\n \\ the job is working remotely, please wait status=done - job_id=3832148009444147886 7886 \n\nquery done succesfully!\n2\n008900250010.001,008900330010.001,008900420010.001,008900430010.001,008900440010.001,008900450010.001,008900460010.001,008900470010.001,008900500010.001,008900510010.001,008900520010.001,008900530010.001,008900540010.001,008900550010.001,008900560010.001,008900570010.001,008900580010.001,008900590010.001,009000010010.001,009000020010.001,009000030010.001,009000040010.001,009000050010.001,009000110010.001,009000120010.001,009000130010.001,020700040010.001,020700050010.001,020700060010.001,020700070010.001,020700110010.001,020700120010.001,020700130010.001,020700140010.001,020700150010.001,020700410010.001,020700420010.001,020700430010.001,020700440010.001,020700450010.001,020700490010.001,020700500010.001,020700510010.001,020700520010.001,020700530010.001,020700540010.001,022600650010.001,026700000110.001,026700020010.001,026700030010.001,026700040010.001\n- waiting for remote response, please wait run_analysis https://www.astro.unige.ch/cdci/astrooda/dispatch-data\nscw_list 008900250010.001,008900330010.001,008900420010.001,008900430010.001,008900440010.001,008900450010.001,008900460010.001,008900470010.001,008900500010.001,008900510010.001,008900520010.001,008900530010.001,008900540010.001,008900550010.001,008900560010.001,008900570010.001,008900580010.001,008900590010.001,009000010010.001,009000020010.001,009000030010.001,009000040010.001,009000050010.001,009000110010.001,009000120010.001,009000130010.001,020700040010.001,020700050010.001,020700060010.001,020700070010.001,020700110010.001,020700120010.001,020700130010.001,020700140010.001,020700150010.001,020700410010.001,020700420010.001,020700430010.001,020700440010.001,020700450010.001,020700490010.001,020700500010.001,020700510010.001,020700520010.001,020700530010.001,020700540010.001,022600650010.001,026700000110.001,026700020010.001,026700030010.001,026700040010.001\nquery_type Real\nosa_version OSA10.2\nRA 194.046527\nDEC -5.789314\nselected_catalog {\"cat_frame\": \"fk5\", \"cat_coord_units\": \"deg\", \"cat_column_list\": [[1, 30, 50, 112, 0], [\"3C 273\", \"IGR J12391-1612\", \"NGC 4593\", \"SWIFT J1238.6+0928\", \"3C 279\"], [12.098819732666016, 4.082015037536621, 13.66150951385498, 4.295688629150391, 0.0], [187.2660675048828, 189.77621459960938, 189.91522216796875, 189.67724609375, 194.046527], [2.045917510986328, -16.179750442504883, -5.358253479003906, 9.474955558776855, -5.789314], [-32768, -32768, -32768, -32768, 0], [2, 2, 2, 1, 2], [0, 0, 0, 0, 0], [0.0002800000074785203, 0.00016999999934341758, 0.0002800000074785203, 0.07566666603088379, 0.0]], \"cat_column_names\": [\"meta_ID\", \"src_names\", \"significance\", \"ra\", \"dec\", \"NEW_SOURCE\", \"ISGRI_FLAG\", \"FLAG\", \"ERR_RAD\"], \"cat_column_descr\": [[\"meta_ID\", \"<i8\"], [\"src_names\", \"<U18\"], [\"significance\", \"<f8\"], [\"ra\", \"<f8\"], [\"dec\", \"<f8\"], [\"NEW_SOURCE\", \"<i8\"], [\"ISGRI_FLAG\", \"<i8\"], [\"FLAG\", \"<i8\"], [\"ERR_RAD\", \"<f8\"]], \"cat_lat_name\": \"dec\", \"cat_lon_name\": \"ra\"}\ninstrument isgri\nproduct_type isgri_spectrum\noff_line (False,)\nquery_status ('new',)\nverbose (False,)\nsession_id 2QP4GBGY4TT5S3SJ\ndry_run (False,)\napi True\n" ], [ "d=spectrum_results[0]\nfor ID,s in enumerate(d._p_list):\n if (s.meta_data['src_name']==source_name):\n if(s.meta_data['product']=='isgri_spectrum'):\n ID_spec=ID\n if(s.meta_data['product']=='isgri_arf'):\n ID_arf=ID\n if(s.meta_data['product']=='isgri_rmf'):\n ID_rmf=ID\n\nprint(ID_spec, ID_arf, ID_rmf)\n ", "3 4 5\n" ], [ "d=spectrum_results[0]\nspec=d._p_list[ID_spec].data_unit[1].data\narf=d._p_list[ID_arf].data_unit[1].data\nrmf=d._p_list[ID_rmf].data_unit[2].data\nch=spec['CHANNEL']\nrate=spec['RATE']*0.\nerr=spec['STAT_ERR']*0.\nsyst=spec['SYS_ERR']*0.\nrate.fill(0)\nerr.fill(0)\nsyst.fill(0)\nqual=spec['QUALITY']\nmatrix=rmf['MATRIX']*0.\nspecresp=arf['SPECRESP']*0.\ntot_expos=0.\ncorr_expos=np.zeros(len(rate))\nprint(len(rate))\nfor k in range(len(scw_lists_osa10)):\n d=spectrum_results[k]\n spec=d._p_list[ID_spec].data_unit[1].data\n arf=d._p_list[ID_arf].data_unit[1].data\n rmf=d._p_list[ID_rmf].data_unit[2].data\n expos=d._p_list[0].data_unit[1].header['EXPOSURE']\n tot_expos=tot_expos+expos\n print(k,expos)\n for j in range(len(rate)):\n if(spec['QUALITY'][j]==0): \n rate[j]=rate[j]+spec['RATE'][j]/(spec['STAT_ERR'][j])**2\n err[j]=err[j]+1./(spec['STAT_ERR'][j])**2\n syst[j]=syst[j]+(spec['SYS_ERR'][j])**2*expos\n corr_expos[j]=corr_expos[j]+expos\n matrix=matrix+rmf['MATRIX']*expos\n specresp=specresp+arf['SPECRESP']*expos\n\nfor i in range(len(rate)):\n if err[i]>0.:\n rate[i]=rate[i]/err[i]\n err[i]=1./sqrt(err[i])\nmatrix=matrix/tot_expos\nspecresp=specresp/tot_expos\nsyst=sqrt(syst/(corr_expos+1.))\nprint('Total exposure:',tot_expos)", "62\n0 115628.4045003584\n1 90360.35488457941\n2 66237.00101908497\n3 28053.87997326724\n4 42416.37498623534\n5 54899.01417923988\n6 72681.13209957794\n7 84306.25795508562\n8 91171.40521538183\n9 67297.86490586787\n10 50185.31346775457\n11 80500.70525250556\n12 44682.28527407059\n13 71992.67794999375\n14 69651.39811800097\n15 61578.40132774157\n16 8651.617763955479\n17 4634.80690293361\n18 2367.31019269395\n19 108753.4897604977\n20 12024.0962724554\n21 10817.12965649085\nTotal exposure: 1238890.9216577725\n" ], [ "print(rate)\nprint(err)", "[ 0.00000000e+00 -1.91108733e-02 -4.74805431e-03 -8.60083848e-03\n 2.89484225e-02 5.46972780e-03 3.97485914e-03 1.83909703e-02\n 1.79907177e-02 1.07161747e-02 1.57599468e-02 1.08829625e-02\n 7.29459012e-03 1.40787875e-02 2.61543579e-02 1.29586281e-02\n 1.04378965e-02 1.47986431e-02 3.07247937e-02 2.76438203e-02\n 1.55900978e-02 2.68101525e-02 2.20132917e-02 1.55036217e-02\n 2.15381514e-02 2.07527503e-02 1.58636644e-02 2.47534411e-03\n 1.27718551e-02 4.38565155e-03 2.54256045e-03 1.35317794e-03\n 1.20742992e-02 1.82996262e-02 1.09611489e-02 2.20161825e-02\n 1.40028633e-02 1.48153109e-02 1.11153908e-02 6.20406447e-03\n -3.77116655e-03 7.94668216e-03 4.36933571e-03 -4.14732331e-03\n -2.96756392e-03 6.25281420e-04 3.96637991e-03 3.72313219e-03\n -7.48174358e-03 -7.23820413e-03 -1.32521137e-03 -6.03199405e-05\n 8.40168097e-04 5.49091585e-03 4.69506858e-03 7.58503389e-04\n 3.12829099e-04 -2.72279885e-03 -1.93422602e-03 -9.32901818e-03\n 6.12873537e-03 -2.72820634e-03]\n[0. 0.0695955 0.03435365 0.020544 0.01424723 0.01093074\n 0.00893867 0.00767821 0.00696467 0.00649837 0.00620647 0.00591155\n 0.00568408 0.00543641 0.00699879 0.00603061 0.00526691 0.00482192\n 0.00621557 0.00566362 0.00537316 0.00522471 0.00532552 0.00598756\n 0.00701571 0.00630916 0.00543909 0.00526548 0.00505263 0.00453153\n 0.00410468 0.00426973 0.00431794 0.00425787 0.00484556 0.00446172\n 0.00417033 0.00394553 0.00378355 0.00364721 0.00356687 0.00352672\n 0.0035033 0.00346465 0.0034068 0.00333141 0.00327024 0.00321747\n 0.00441959 0.00424585 0.00403846 0.00384813 0.00368388 0.00355565\n 0.00339815 0.00321211 0.003678 0.00410753 0.00521346 0.005994\n 0.00657711 0.00488587]\n" ], [ "d._p_list[ID_spec].data_unit[1].data['RATE']=rate\nd._p_list[ID_spec].data_unit[1].data['STAT_ERR']=err\nd._p_list[ID_rmf].data_unit[2].data['MATRIX']=matrix\nd._p_list[ID_arf].data_unit[1].data['SPECRESP']=specresp", "_____no_output_____" ], [ "name=source_name.replace(\" \", \"\")\nspecname=name+'_spectrum_osa10.fits'\narfname=name+'_arf_osa10.fits.gz'\nrmfname=name+'_rmf_osa10.fits.gz'\ndata._p_list[ID_spec].write_fits_file(specname)\ndata._p_list[ID_arf].write_fits_file(arfname)\ndata._p_list[ID_rmf].write_fits_file(rmfname)", "WARNING: VerifyWarning: Keyword name 'instrument' is greater than 8 characters or contains characters not allowed by the FITS standard; a HIERARCH card will be created. [astropy.io.fits.card]\nWARNING: VerifyWarning: Keyword name 'osa_version' is greater than 8 characters or contains characters not allowed by the FITS standard; a HIERARCH card will be created. [astropy.io.fits.card]\nWARNING: VerifyWarning: Keyword name 'product_type' is greater than 8 characters or contains characters not allowed by the FITS standard; a HIERARCH card will be created. [astropy.io.fits.card]\nWARNING: VerifyWarning: Keyword name 'query_status' is greater than 8 characters or contains characters not allowed by the FITS standard; a HIERARCH card will be created. [astropy.io.fits.card]\nWARNING: VerifyWarning: Keyword name 'query_type' is greater than 8 characters or contains characters not allowed by the FITS standard; a HIERARCH card will be created. [astropy.io.fits.card]\nWARNING: VerifyWarning: Keyword name 'session_id' is greater than 8 characters or contains characters not allowed by the FITS standard; a HIERARCH card will be created. [astropy.io.fits.card]\n" ], [ "hdul = fits.open(specname, mode='update')\nhdr=hdul[1].header\nhdr.set('EXPOSURE', tot_expos)\nhdul.close()", "_____no_output_____" ], [ "!./spectrum_fit_osa10.sh $name $rebin", "\n\t\tXSPEC version: 12.9.0n\n\tBuild Date/Time: Tue Nov 8 18:03:34 2016\n\nXSPEC12>statistic chi\nDefault fit statistic is set to: Chi-Squared\n This will apply to all current and newly loaded spectra.\nXSPEC12>data 3C279_spectrum_osa10.fits\n***Warning: POISSERR keyword is missing or of wrong format, assuming FALSE.\n\n1 spectrum in use\n \nSpectral Data File: 3C279_spectrum_osa10.fits Spectrum 1\nNet count rate (cts/s) for Spectrum:1 4.900e-01 +/- 9.055e-02\n Assigned to Data Group 1 and Plot Group 1\n Noticed Channels: 1-62\n Telescope: INTEGRAL Instrument: IBIS Channel Type: PI\n Exposure Time: 1.239e+06 sec\n Using fit statistic: chi\n Using test statistic: chi\n No response loaded.\n\n***Warning! One or more spectra are missing responses,\n and are not suitable for fit.\nXSPEC12>response 3C279_rmf_osa10.fits.gz\nWarning: RMF CHANTYPE keyword (PHA) is not consistent with that from spectrum (PI)\nResponse successfully loaded.\nXSPEC12>arf 3C279_arf_osa10.fits.gz\nArf successfully loaded.\nXSPEC12>ignore bad\n\nignore: 6 channels ignored from source number 1\nXSPEC12>ignore 300.-**\n 5 channels (58-62) ignored in spectrum # 1\n\nXSPEC12>model powerlaw\n\nInput parameter value, delta, min, bot, top, and max values for ...\n 1 0.01( 0.01) -3 -2 9 10\n 10erlaw:PhoIndex> 1. 0.01 -3 -2 9 \n 1 0.01( 0.01) 0 0 1e+20 1e+24\n1e+24erlaw:norm> 0.01 0.01 0 0 1e+20 \n\n========================================================================\nModel powerlaw<1> Source No.: 1 Active/On\nModel Model Component Parameter Unit Value\n par comp\n 1 1 powerlaw PhoIndex 1.00000 +/- 0.0 \n 2 1 powerlaw norm 1.00000E-02 +/- 0.0 \n________________________________________________________________________\n\n\nFit statistic : Chi-Squared = 478951.9 using 51 PHA bins.\n\nTest statistic : Chi-Squared = 478951.9 using 51 PHA bins.\n Reduced chi-squared = 9774.529 for 49 degrees of freedom \n Null hypothesis probability = 0.000000e+00\n Current data and model not fit yet.\nXSPEC12>fit 1000\n Parameters\nChi-Squared |beta|/N Lvl 1:PhoIndex 2:norm\n78.824 14.6109 -1 1.03285 0.000248154\n78.7438 10879.1 -2 1.19254 0.000415258\n78.2251 39615.8 -3 1.33887 0.000733991\n71.434 23648.9 -4 1.30796 0.000789235\n71.3746 2376.55 -5 1.31686 0.000805472\n71.3745 11.4964 -6 1.31586 0.000802374\n==============================\n Variances and Principal Axes\n 1 2 \n 1.2961E-02| -1.0000 -0.0032 \n 2.5608E-09| 0.0032 -1.0000 \n------------------------------\n\n========================\n Covariance Matrix\n 1 2 \n 1.296e-02 4.092e-05\n 4.092e-05 1.318e-07\n------------------------\n\n========================================================================\nModel powerlaw<1> Source No.: 1 Active/On\nModel Model Component Parameter Unit Value\n par comp\n 1 1 powerlaw PhoIndex 1.31586 +/- 0.113847 \n 2 1 powerlaw norm 8.02374E-04 +/- 3.62992E-04 \n________________________________________________________________________\n\n\nFit statistic : Chi-Squared = 71.37 using 51 PHA bins.\n\nTest statistic : Chi-Squared = 71.37 using 51 PHA bins.\n Reduced chi-squared = 1.457 for 49 degrees of freedom \n Null hypothesis probability = 2.012080e-02\nXSPEC12>setplot rebin 10 10\nXSPEC12>plot eeufs\nXSPEC12>iplot\nPLT> wdata 3C279_spectrum_osa10.txt\nPLT> XSPEC12>\n XSPEC: quit\n" ], [ "spectrum_results1=[]\nfor i in range(len(scw_lists_osa11)):\n print(i)\n data=disp.get_product(instrument='isgri',\n product='isgri_spectrum',\n scw_list=scw_lists_osa11[i],\n query_type='Real',\n osa_version='OSA11.0',\n RA=ra,\n DEC=dec,\n product_type='Real',\n selected_catalog=api_cat)\n spectrum_results1.append(data)", "0\n169500140010.001,169500150010.001,169500210010.001,169500220010.001,169500230010.001,169500240010.001,169500250010.001,169500260010.001,169500270010.001,169900840010.001,169900850010.001,175400370010.001,176000060010.001,176100050010.001,176200040010.001,176200050010.001,176200450010.001,176300030010.001,176300040010.001,176400060010.001,176400460010.001,176400470010.001,176500360010.001,176500450010.001,176500460010.001,182000220010.001,182000310010.001,182100250010.001,182100340010.001,182200220010.001,182200230010.001,182200240010.001,182200320010.001,182300600010.001,182400030010.001,182400430010.001,182400440010.001,182800300010.001,182800400010.001,183200120010.001,183200130010.001,183200190010.001,183200200010.001,183200210010.001,183200220010.001,183200230010.001,183200240010.001,183200250010.001,183200290010.001,183200300010.001\n- waiting for remote response, please wait run_analysis https://www.astro.unige.ch/cdci/astrooda/dispatch-data\nscw_list 169500140010.001,169500150010.001,169500210010.001,169500220010.001,169500230010.001,169500240010.001,169500250010.001,169500260010.001,169500270010.001,169900840010.001,169900850010.001,175400370010.001,176000060010.001,176100050010.001,176200040010.001,176200050010.001,176200450010.001,176300030010.001,176300040010.001,176400060010.001,176400460010.001,176400470010.001,176500360010.001,176500450010.001,176500460010.001,182000220010.001,182000310010.001,182100250010.001,182100340010.001,182200220010.001,182200230010.001,182200240010.001,182200320010.001,182300600010.001,182400030010.001,182400430010.001,182400440010.001,182800300010.001,182800400010.001,183200120010.001,183200130010.001,183200190010.001,183200200010.001,183200210010.001,183200220010.001,183200230010.001,183200240010.001,183200250010.001,183200290010.001,183200300010.001\nquery_type Real\nosa_version OSA11.0\nRA 194.046527\nDEC -5.789314\nselected_catalog {\"cat_frame\": \"fk5\", \"cat_coord_units\": \"deg\", \"cat_column_list\": [[1, 30, 50, 112, 0], [\"3C 273\", \"IGR J12391-1612\", \"NGC 4593\", \"SWIFT J1238.6+0928\", \"3C 279\"], [12.098819732666016, 4.082015037536621, 13.66150951385498, 4.295688629150391, 0.0], [187.2660675048828, 189.77621459960938, 189.91522216796875, 189.67724609375, 194.046527], [2.045917510986328, -16.179750442504883, -5.358253479003906, 9.474955558776855, -5.789314], [-32768, -32768, -32768, -32768, 0], [2, 2, 2, 1, 2], [0, 0, 0, 0, 0], [0.0002800000074785203, 0.00016999999934341758, 0.0002800000074785203, 0.07566666603088379, 0.0]], \"cat_column_names\": [\"meta_ID\", \"src_names\", \"significance\", \"ra\", \"dec\", \"NEW_SOURCE\", \"ISGRI_FLAG\", \"FLAG\", \"ERR_RAD\"], \"cat_column_descr\": [[\"meta_ID\", \"<i8\"], [\"src_names\", \"<U18\"], [\"significance\", \"<f8\"], [\"ra\", \"<f8\"], [\"dec\", \"<f8\"], [\"NEW_SOURCE\", \"<i8\"], [\"ISGRI_FLAG\", \"<i8\"], [\"FLAG\", \"<i8\"], [\"ERR_RAD\", \"<f8\"]], \"cat_lat_name\": \"dec\", \"cat_lon_name\": \"ra\"}\ninstrument isgri\nproduct_type isgri_spectrum\noff_line (False,)\nquery_status ('new',)\nverbose (False,)\nsession_id FT1WC6B0CVLO15D4\ndry_run (False,)\napi True\nthe job has been submitted on the remote server\n / the job is working remotely, please wait status=done - job_id=-98497454171466425 6425 \n\nquery done succesfully!\n1\n183200310010.001,183200320010.001,183200330010.001,183200340010.001,183200400010.001,183200410010.001,184500030010.001,184500040010.001,184500430010.001,184600020010.001,184600100010.001,184600110010.001,184600120010.001,185000310010.001,189200400010.001,189200410010.001,189200420010.001,189200490010.001,189200500010.001,189300020010.001,189300410010.001,189300420010.001,189400030010.001,189400040010.001,189400050010.001,189400450010.001,189500070010.001,189500480010.001,189600100010.001,189700030010.001,189700040010.001,189700130010.001,189800060010.001,189800070010.001,189800080010.001,189800160010.001,189900080010.001,189900090010.001,189900170010.001,189900180010.001,189900190010.001,190000110010.001,190000120010.001,190000130010.001,190000210010.001,190100130010.001,190100140010.001\n- waiting for remote response, please wait run_analysis https://www.astro.unige.ch/cdci/astrooda/dispatch-data\nscw_list 183200310010.001,183200320010.001,183200330010.001,183200340010.001,183200400010.001,183200410010.001,184500030010.001,184500040010.001,184500430010.001,184600020010.001,184600100010.001,184600110010.001,184600120010.001,185000310010.001,189200400010.001,189200410010.001,189200420010.001,189200490010.001,189200500010.001,189300020010.001,189300410010.001,189300420010.001,189400030010.001,189400040010.001,189400050010.001,189400450010.001,189500070010.001,189500480010.001,189600100010.001,189700030010.001,189700040010.001,189700130010.001,189800060010.001,189800070010.001,189800080010.001,189800160010.001,189900080010.001,189900090010.001,189900170010.001,189900180010.001,189900190010.001,190000110010.001,190000120010.001,190000130010.001,190000210010.001,190100130010.001,190100140010.001\nquery_type Real\nosa_version OSA11.0\nRA 194.046527\nDEC -5.789314\nselected_catalog {\"cat_frame\": \"fk5\", \"cat_coord_units\": \"deg\", \"cat_column_list\": [[1, 30, 50, 112, 0], [\"3C 273\", \"IGR J12391-1612\", \"NGC 4593\", \"SWIFT J1238.6+0928\", \"3C 279\"], [12.098819732666016, 4.082015037536621, 13.66150951385498, 4.295688629150391, 0.0], [187.2660675048828, 189.77621459960938, 189.91522216796875, 189.67724609375, 194.046527], [2.045917510986328, -16.179750442504883, -5.358253479003906, 9.474955558776855, -5.789314], [-32768, -32768, -32768, -32768, 0], [2, 2, 2, 1, 2], [0, 0, 0, 0, 0], [0.0002800000074785203, 0.00016999999934341758, 0.0002800000074785203, 0.07566666603088379, 0.0]], \"cat_column_names\": [\"meta_ID\", \"src_names\", \"significance\", \"ra\", \"dec\", \"NEW_SOURCE\", \"ISGRI_FLAG\", \"FLAG\", \"ERR_RAD\"], \"cat_column_descr\": [[\"meta_ID\", \"<i8\"], [\"src_names\", \"<U18\"], [\"significance\", \"<f8\"], [\"ra\", \"<f8\"], [\"dec\", \"<f8\"], [\"NEW_SOURCE\", \"<i8\"], [\"ISGRI_FLAG\", \"<i8\"], [\"FLAG\", \"<i8\"], [\"ERR_RAD\", \"<f8\"]], \"cat_lat_name\": \"dec\", \"cat_lon_name\": \"ra\"}\ninstrument isgri\nproduct_type isgri_spectrum\noff_line (False,)\nquery_status ('new',)\nverbose (False,)\nsession_id M822KTIZCGI4I5NY\ndry_run (False,)\napi True\nthe job has been submitted on the remote server\n \\ the job is working remotely, please wait status=done - job_id=8843659157675212506 2506 \n\nquery done succesfully!\n" ], [ "d=spectrum_results1[0]\nfor ID,s in enumerate(d._p_list):\n if (s.meta_data['src_name']==source_name):\n if(s.meta_data['product']=='isgri_spectrum'):\n ID_spec=ID\n if(s.meta_data['product']=='isgri_arf'):\n ID_arf=ID\n if(s.meta_data['product']=='isgri_rmf'):\n ID_rmf=ID\n\nprint(ID_spec, ID_arf, ID_rmf)\n\n", "3 4 5\n" ], [ "d=spectrum_results1[0]\nspec=d._p_list[ID_spec].data_unit[1].data\narf=d._p_list[ID_arf].data_unit[1].data\nrmf=d._p_list[ID_rmf].data_unit[2].data\nch=spec['CHANNEL']\nrate=spec['RATE']*0.\nerr=spec['STAT_ERR']*0.\nsyst=spec['SYS_ERR']*0.\nrate.fill(0)\nerr.fill(0)\nsyst.fill(0)\nqual=spec['QUALITY']\nmatrix=rmf['MATRIX']*0.\nspecresp=arf['SPECRESP']*0.\ntot_expos=0.\ncorr_expos=np.zeros(len(rate))\nprint(len(rate))\nfor k in range(len(scw_lists_osa11)):\n d=spectrum_results1[k]\n spec=d._p_list[ID_spec].data_unit[1].data\n arf=d._p_list[ID_arf].data_unit[1].data\n rmf=d._p_list[ID_rmf].data_unit[2].data\n expos=d._p_list[0].data_unit[1].header['EXPOSURE']\n tot_expos=tot_expos+expos\n print(k,expos)\n for j in range(len(rate)):\n if(spec['QUALITY'][j]==0): \n rate[j]=rate[j]+spec['RATE'][j]/(spec['STAT_ERR'][j])**2\n err[j]=err[j]+1./(spec['STAT_ERR'][j])**2\n syst[j]=syst[j]+(spec['SYS_ERR'][j])**2*expos\n corr_expos[j]=corr_expos[j]+expos\n matrix=matrix+rmf['MATRIX']*expos\n specresp=specresp+arf['SPECRESP']*expos\n\nfor i in range(len(rate)):\n if err[i]>0.:\n rate[i]=rate[i]/err[i]\n err[i]=1./sqrt(err[i])\nmatrix=matrix/tot_expos\nspecresp=specresp/tot_expos\nsyst=sqrt(syst/(corr_expos+1.))\nprint('Total exposure:',tot_expos)", "256\n0 26026.8547870585\n1 8087.67341090079\nTotal exposure: 34114.52819795929\n" ], [ "d._p_list[ID_spec].data_unit[1].data['RATE']=rate\nd._p_list[ID_spec].data_unit[1].data['STAT_ERR']=err\nd._p_list[ID_rmf].data_unit[2].data['MATRIX']=matrix\nd._p_list[ID_arf].data_unit[1].data['SPECRESP']=specresp", "_____no_output_____" ], [ "name=source_name.replace(\" \", \"\")\nspecname=name+'_spectrum_osa11.fits'\narfname=name+'_arf_osa11.fits.gz'\nrmfname=name+'_rmf_osa11.fits.gz'\ndata._p_list[ID_spec].write_fits_file(specname)\ndata._p_list[ID_arf].write_fits_file(arfname)\ndata._p_list[ID_rmf].write_fits_file(rmfname)", "_____no_output_____" ], [ "hdul = fits.open(specname, mode='update')\nhdr=hdul[1].header\nhdr.set('EXPOSURE', tot_expos)\nhdul.close()", "_____no_output_____" ], [ "!./spectrum_fit_osa11.sh $name $rebin", "\n\t\tXSPEC version: 12.9.0n\n\tBuild Date/Time: Tue Nov 8 18:03:34 2016\n\nXSPEC12>statistic chi\nDefault fit statistic is set to: Chi-Squared\n This will apply to all current and newly loaded spectra.\nXSPEC12>data 3C279_spectrum_osa11.fits\n***Warning: POISSERR keyword is missing or of wrong format, assuming FALSE.\n\n1 spectrum in use\n \nSpectral Data File: 3C279_spectrum_osa11.fits Spectrum 1\nNet count rate (cts/s) for Spectrum:1 2.216e-01 +/- 1.530e-01\n Assigned to Data Group 1 and Plot Group 1\n Noticed Channels: 1-256\n Telescope: INTEGRAL Instrument: IBIS Channel Type: PI\n Exposure Time: 3.411e+04 sec\n Using fit statistic: chi\n Using test statistic: chi\n No response loaded.\n\n***Warning! One or more spectra are missing responses,\n and are not suitable for fit.\nXSPEC12>response 3C279_rmf_osa11.fits.gz\nResponse successfully loaded.\nXSPEC12>arf 3C279_arf_osa11.fits.gz\nArf successfully loaded.\nXSPEC12>ignore bad\n\nignore: 25 channels ignored from source number 1\nXSPEC12>ignore 300.-**\n 128 channels (129-256) ignored in spectrum # 1\n\nXSPEC12>model powerlaw\n\nInput parameter value, delta, min, bot, top, and max values for ...\n 1 0.01( 0.01) -3 -2 9 10\n 10erlaw:PhoIndex> 1. 0.01 -3 -2 9 \n 1 0.01( 0.01) 0 0 1e+20 1e+24\n1e+24erlaw:norm> 0.01 0.01 0 0 1e+20 \n\n========================================================================\nModel powerlaw<1> Source No.: 1 Active/On\nModel Model Component Parameter Unit Value\n par comp\n 1 1 powerlaw PhoIndex 1.00000 +/- 0.0 \n 2 1 powerlaw norm 1.00000E-02 +/- 0.0 \n________________________________________________________________________\n\n\nFit statistic : Chi-Squared = 26502.26 using 106 PHA bins.\n\nTest statistic : Chi-Squared = 26502.26 using 106 PHA bins.\n Reduced chi-squared = 254.8294 for 104 degrees of freedom \n Null hypothesis probability = 0.000000e+00\n Current data and model not fit yet.\nXSPEC12>fit 1000\n Parameters\nChi-Squared |beta|/N Lvl 1:PhoIndex 2:norm\n116.989 2.33136 -1 1.15343 0.000274912\n115.995 3073.36 -1 1.29865 0.000500267\n114.786 2184.51 -1 1.42512 0.000910125\n113.645 1240.59 -1 1.53466 0.00155192\n112.683 682.875 -1 1.63210 0.00246531\n111.88 389.471 -1 1.72076 0.00370055\n111.194 234.238 -1 1.80254 0.00532162\n110.597 148.006 -1 1.87866 0.00740322\n110.068 97.4112 -1 1.94998 0.0100289\n109.595 66.2761 -1 2.01715 0.0132907\n109.167 46.352 -1 2.08070 0.0172897\n108.778 33.184 -1 2.14105 0.0221363\n108.422 24.2413 -1 2.19855 0.0279507\n108.095 18.0245 -1 2.25349 0.0348639\n107.792 13.6133 -1 2.30611 0.0430177\n107.512 10.4262 -1 2.35663 0.0525655\n107.25 8.08596 -1 2.40524 0.0636732\n107.006 6.34243 -1 2.45208 0.0765190\n106.777 5.02623 -1 2.49730 0.0912946\n106.562 4.02065 -1 2.54103 0.108205\n106.36 3.24393 -1 2.58336 0.127472\n106.168 2.63793 -1 2.62441 0.149328\n105.988 2.16075 -1 2.66424 0.174026\n105.816 1.7818 -1 2.70295 0.201830\n105.654 1.47852 -1 2.74059 0.233026\n105.499 1.23405 -1 2.77725 0.267914\n105.352 1.03572 -1 2.81296 0.306812\n105.211 0.873875 -1 2.84778 0.350058\n105.077 0.741135 -1 2.88176 0.398008\n104.949 0.631792 -1 2.91495 0.451036\n104.826 0.541413 -1 2.94738 0.509540\n104.708 0.466526 -1 2.97909 0.573935\n104.595 0.4044 -1 3.01012 0.644658\n104.487 0.352869 -1 3.04049 0.722168\n104.382 0.310204 -1 3.07023 0.806947\n104.282 0.27501 -1 3.09938 0.899497\n104.185 0.246145 -1 3.12795 1.00035\n104.092 0.222663 -1 3.15597 1.11004\n104.003 0.203756 -1 3.18346 1.22916\n103.916 0.188726 -1 3.21044 1.35830\n103.833 0.176955 -1 3.23693 1.49809\n103.752 0.167894 -1 3.26295 1.64916\n103.674 0.161055 -1 3.28852 1.81219\n103.598 0.156008 -1 3.31364 1.98789\n103.525 0.152384 -1 3.33834 2.17697\n103.455 0.149872 -1 3.36263 2.38019\n103.386 0.148214 -1 3.38652 2.59833\n103.32 0.147201 -1 3.41003 2.83217\n103.255 0.146667 -1 3.43316 3.08257\n103.193 0.146481 -1 3.45593 3.35037\n103.132 0.146543 -1 3.47835 3.63645\n103.073 0.146775 -1 3.50043 3.94173\n103.016 0.147117 -1 3.52218 4.26715\n102.961 0.147525 -1 3.54360 4.61366\n102.907 0.147965 -1 3.56471 4.98227\n102.855 0.148413 -1 3.58551 5.37399\n102.804 0.14885 -1 3.60602 5.78987\n102.754 0.149263 -1 3.62623 6.23097\n102.706 0.149643 -1 3.64617 6.69841\n102.659 0.149983 -1 3.66582 7.19331\n102.613 0.150279 -1 3.68521 7.71681\n102.569 0.15053 -1 3.70434 8.27010\n102.526 0.150732 -1 3.72320 8.85439\n102.484 0.150887 -1 3.74182 9.47090\n102.442 0.150994 -1 3.76019 10.1209\n102.402 0.151054 -1 3.77833 10.8056\n102.363 0.151069 -1 3.79622 11.5265\n102.325 0.151039 -1 3.81389 12.2847\n102.288 0.150966 -1 3.83134 13.0816\n102.252 0.150853 -1 3.84856 13.9187\n102.217 0.150701 -1 3.86557 14.7974\n102.182 0.150511 -1 3.88236 15.7189\n102.149 0.150285 -1 3.89895 16.6849\n102.116 0.150026 -1 3.91534 17.6967\n102.112 0.149735 -2 4.06364 27.0748\n==============================\n Variances and Principal Axes\n 1 2 \n 3.3554E-03| -0.9999 0.0159 \n 5.3526E+03| -0.0159 -0.9999 \n------------------------------\n\n========================\n Covariance Matrix\n 1 2 \n 1.350e+00 8.490e+01\n 8.490e+01 5.351e+03\n------------------------\n\n========================================================================\nModel powerlaw<1> Source No.: 1 Active/On\nModel Model Component Parameter Unit Value\n par comp\n 1 1 powerlaw PhoIndex 4.06364 +/- 1.16199 \n 2 1 powerlaw norm 27.0748 +/- 73.1522 \n________________________________________________________________________\n\n\nFit statistic : Chi-Squared = 102.11 using 106 PHA bins.\n\nTest statistic : Chi-Squared = 102.11 using 106 PHA bins.\n Reduced chi-squared = 0.98185 for 104 degrees of freedom \n Null hypothesis probability = 5.340167e-01\nXSPEC12>setplot rebin 10 10\nXSPEC12>plot eeufs\nXSPEC12>iplot\nPLT> wdata 3C279_spectrum_osa11.txt\nPLT> XSPEC12>\n XSPEC: quit\n" ], [ "data=disp.get_product(instrument='isgri',\n product='isgri_spectrum',\n T1='2015-06-15T15:56:45',\n T2='2015-06-16T06:13:10',\n query_type='Real',\n osa_version='OSA10.2',\n RA=ra,\n DEC=dec,\n detection_threshold=5.0,\n radius=15.,\n product_type='Real',\n selected_catalog=api_cat)", "- waiting for remote response, please wait run_analysis https://www.astro.unige.ch/cdci/astrooda/dispatch-data\nT1 2015-06-15T15:56:45\nT2 2015-06-16T06:13:10\nquery_type Real\nosa_version OSA10.2\nRA 194.046527\nDEC -5.789314\ndetection_threshold 5.0\nradius 15.0\nselected_catalog {\"cat_frame\": \"fk5\", \"cat_coord_units\": \"deg\", \"cat_column_list\": [[1, 30, 50, 112, 0], [\"3C 273\", \"IGR J12391-1612\", \"NGC 4593\", \"SWIFT J1238.6+0928\", \"3C 279\"], [12.098819732666016, 4.082015037536621, 13.66150951385498, 4.295688629150391, 0.0], [187.2660675048828, 189.77621459960938, 189.91522216796875, 189.67724609375, 194.046527], [2.045917510986328, -16.179750442504883, -5.358253479003906, 9.474955558776855, -5.789314], [-32768, -32768, -32768, -32768, 0], [2, 2, 2, 1, 2], [0, 0, 0, 0, 0], [0.0002800000074785203, 0.00016999999934341758, 0.0002800000074785203, 0.07566666603088379, 0.0]], \"cat_column_names\": [\"meta_ID\", \"src_names\", \"significance\", \"ra\", \"dec\", \"NEW_SOURCE\", \"ISGRI_FLAG\", \"FLAG\", \"ERR_RAD\"], \"cat_column_descr\": [[\"meta_ID\", \"<i8\"], [\"src_names\", \"<U18\"], [\"significance\", \"<f8\"], [\"ra\", \"<f8\"], [\"dec\", \"<f8\"], [\"NEW_SOURCE\", \"<i8\"], [\"ISGRI_FLAG\", \"<i8\"], [\"FLAG\", \"<i8\"], [\"ERR_RAD\", \"<f8\"]], \"cat_lat_name\": \"dec\", \"cat_lon_name\": \"ra\"}\ninstrument isgri\nproduct_type isgri_spectrum\noff_line (False,)\nquery_status ('new',)\nverbose (False,)\nsession_id 1SWWY8TILU3Z3XFF\ndry_run (False,)\napi True\nthe job has been submitted on the remote server\n / the job is working remotely, please wait status=done - job_id=6490951792990391550 1550 \n\nquery done succesfully!\n" ], [ "data._p_list[0].write_fits_file(name+'_flare_spectrum_osa10.fits')\ndata._p_list[1].write_fits_file(name+'_flare_arf_osa10.fits.gz')\ndata._p_list[2].write_fits_file(name+'_flare_rmf_osa10.fits.gz')", "WARNING: VerifyWarning: Keyword name 'detection_threshold' is greater than 8 characters or contains characters not allowed by the FITS standard; a HIERARCH card will be created. [astropy.io.fits.card]\n" ], [ "name1=name+'_flare'\n!./spectrum_fit_osa10.sh $name1 $rebin", "\n\t\tXSPEC version: 12.9.0n\n\tBuild Date/Time: Tue Nov 8 18:03:34 2016\n\nXSPEC12>statistic chi\nDefault fit statistic is set to: Chi-Squared\n This will apply to all current and newly loaded spectra.\nXSPEC12>data 3C279_flare_spectrum_osa10.fits\n***Warning: POISSERR keyword is missing or of wrong format, assuming FALSE.\n\n1 spectrum in use\n \nSpectral Data File: 3C279_flare_spectrum_osa10.fits Spectrum 1\nNet count rate (cts/s) for Spectrum:1 nan +/- nan \n Assigned to Data Group 1 and Plot Group 1\n Noticed Channels: 1-62\n Telescope: INTEGRAL Instrument: IBIS Channel Type: PI\n Exposure Time: 2.135e+04 sec\n Using fit statistic: chi\n Using test statistic: chi\n No response loaded.\n\n***Warning! One or more spectra are missing responses,\n and are not suitable for fit.\nXSPEC12>response 3C279_flare_rmf_osa10.fits.gz\nWarning: RMF CHANTYPE keyword (PHA) is not consistent with that from spectrum (PI)\nResponse successfully loaded.\nXSPEC12>arf 3C279_flare_arf_osa10.fits.gz\nArf successfully loaded.\nXSPEC12>ignore bad\n\nignore: 6 channels ignored from source number 1\nXSPEC12>ignore 300.-**\n 5 channels (58-62) ignored in spectrum # 1\n\nXSPEC12>model powerlaw\n\nInput parameter value, delta, min, bot, top, and max values for ...\n 1 0.01( 0.01) -3 -2 9 10\n 10erlaw:PhoIndex> 1. 0.01 -3 -2 9 \n 1 0.01( 0.01) 0 0 1e+20 1e+24\n1e+24erlaw:norm> 0.01 0.01 0 0 1e+20 \n\n========================================================================\nModel powerlaw<1> Source No.: 1 Active/On\nModel Model Component Parameter Unit Value\n par comp\n 1 1 powerlaw PhoIndex 1.00000 +/- 0.0 \n 2 1 powerlaw norm 1.00000E-02 +/- 0.0 \n________________________________________________________________________\n\n\nFit statistic : Chi-Squared = 4756.72 using 51 PHA bins.\n\nTest statistic : Chi-Squared = 4756.72 using 51 PHA bins.\n Reduced chi-squared = 97.0759 for 49 degrees of freedom \n Null hypothesis probability = 0.000000e+00\n Current data and model not fit yet.\nXSPEC12>fit 1000\n Parameters\nChi-Squared |beta|/N Lvl 1:PhoIndex 2:norm\n58.6273 2.30987 -1 0.969381 0.000753270\n58.4088 34.0217 -1 0.942119 0.000673677\n58.2313 65.0327 -1 0.917774 0.000609234\n58.0868 92.3293 -1 0.895974 0.000556413\n57.9688 115.611 -1 0.876407 0.000512643\n57.8721 134.835 -1 0.858807 0.000476020\n57.7928 150.135 -1 0.842945 0.000445108\n57.7722 161.762 -2 0.759106 0.000284730\n57.5045 5503.98 -3 0.671853 0.000201620\n57.4109 4682.93 -4 0.684531 0.000222793\n57.4102 373.728 -5 0.682094 0.000221285\n==============================\n Variances and Principal Axes\n 1 2 \n 7.9836E-02| -1.0000 -0.0010 \n 1.1818E-09| 0.0010 -1.0000 \n------------------------------\n\n========================\n Covariance Matrix\n 1 2 \n 7.984e-02 7.832e-05\n 7.832e-05 7.802e-08\n------------------------\n\n========================================================================\nModel powerlaw<1> Source No.: 1 Active/On\nModel Model Component Parameter Unit Value\n par comp\n 1 1 powerlaw PhoIndex 0.682094 +/- 0.282552 \n 2 1 powerlaw norm 2.21285E-04 +/- 2.79314E-04 \n________________________________________________________________________\n\n\nFit statistic : Chi-Squared = 57.41 using 51 PHA bins.\n\nTest statistic : Chi-Squared = 57.41 using 51 PHA bins.\n Reduced chi-squared = 1.172 for 49 degrees of freedom \n Null hypothesis probability = 1.916593e-01\nXSPEC12>setplot rebin 10 10\nXSPEC12>plot eeufs\nXSPEC12>iplot\nPLT> wdata 3C279_flare_spectrum_osa10.txt\nPLT> XSPEC12>\n XSPEC: quit\n" ], [ "plt.figure(figsize=(10,7))\n\nspectrum=np.genfromtxt(name+'_spectrum_osa10.txt',skip_header=3)\nen=spectrum[:,0]\nen_err=spectrum[:,1]\nfl=spectrum[:,2]\nfl_err=spectrum[:,3]\nmo=spectrum[:,4]\nplt.errorbar(en,fl,xerr=en_err,yerr=fl_err,linestyle='none',linewidth=8,color='black',alpha=0.5)\nplt.plot(en,mo,color='black',linewidth=4)\n\n\nspectrum=np.genfromtxt(name+'_flare_spectrum_osa10.txt',skip_header=3)\nen=spectrum[:,0]\nen_err=spectrum[:,1]\nfl=spectrum[:,2]\nfl_err=spectrum[:,3]\nmo=spectrum[:,4]\nplt.errorbar(en,fl,xerr=en_err,yerr=fl_err,linestyle='none',linewidth=8,color='blue',alpha=0.5)\nplt.plot(en,mo,color='blue',linewidth=4,alpha=0.5)\n\n\nplt.tick_params(axis='both', which='major', labelsize=16)\nplt.xscale('log')\nplt.yscale('log')\nplt.ylim(1.e-3,3.e0)\nplt.xlim(15,350)\nplt.xlabel('$E$, keV',fontsize=16)\nplt.ylabel('$E^2F_E$, keV/(cm$^2$s)',fontsize=16)\nplt.savefig(name+'_spectra.pdf',format='pdf',dpi=100)\n\n\n", "_____no_output_____" ], [ "spectrum_3C279=name+'_spectra.pdf'", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Table of Contents\n <p><div class=\"lev1 toc-item\"><a href=\"#Lambda-calcul-implémenté-en-OCaml\" data-toc-modified-id=\"Lambda-calcul-implémenté-en-OCaml-1\"><span class=\"toc-item-num\">1&nbsp;&nbsp;</span>Lambda-calcul implémenté en OCaml</a></div><div class=\"lev2 toc-item\"><a href=\"#Expressions\" data-toc-modified-id=\"Expressions-11\"><span class=\"toc-item-num\">1.1&nbsp;&nbsp;</span>Expressions</a></div><div class=\"lev2 toc-item\"><a href=\"#But-?\" data-toc-modified-id=\"But-?-12\"><span class=\"toc-item-num\">1.2&nbsp;&nbsp;</span>But ?</a></div><div class=\"lev2 toc-item\"><a href=\"#Grammaire\" data-toc-modified-id=\"Grammaire-13\"><span class=\"toc-item-num\">1.3&nbsp;&nbsp;</span>Grammaire</a></div><div class=\"lev2 toc-item\"><a href=\"#L'identité\" data-toc-modified-id=\"L'identité-14\"><span class=\"toc-item-num\">1.4&nbsp;&nbsp;</span>L'identité</a></div><div class=\"lev2 toc-item\"><a href=\"#Conditionnelles\" data-toc-modified-id=\"Conditionnelles-15\"><span class=\"toc-item-num\">1.5&nbsp;&nbsp;</span>Conditionnelles</a></div><div class=\"lev2 toc-item\"><a href=\"#Nombres\" data-toc-modified-id=\"Nombres-16\"><span class=\"toc-item-num\">1.6&nbsp;&nbsp;</span>Nombres</a></div><div class=\"lev2 toc-item\"><a href=\"#Test-d'inégalité\" data-toc-modified-id=\"Test-d'inégalité-17\"><span class=\"toc-item-num\">1.7&nbsp;&nbsp;</span>Test d'inégalité</a></div><div class=\"lev2 toc-item\"><a href=\"#Successeurs\" data-toc-modified-id=\"Successeurs-18\"><span class=\"toc-item-num\">1.8&nbsp;&nbsp;</span>Successeurs</a></div><div class=\"lev2 toc-item\"><a href=\"#Prédecesseurs\" data-toc-modified-id=\"Prédecesseurs-19\"><span class=\"toc-item-num\">1.9&nbsp;&nbsp;</span>Prédecesseurs</a></div><div class=\"lev2 toc-item\"><a href=\"#Addition\" data-toc-modified-id=\"Addition-110\"><span class=\"toc-item-num\">1.10&nbsp;&nbsp;</span>Addition</a></div><div class=\"lev2 toc-item\"><a href=\"#Multiplication\" data-toc-modified-id=\"Multiplication-111\"><span class=\"toc-item-num\">1.11&nbsp;&nbsp;</span>Multiplication</a></div><div class=\"lev2 toc-item\"><a href=\"#Paires\" data-toc-modified-id=\"Paires-112\"><span class=\"toc-item-num\">1.12&nbsp;&nbsp;</span>Paires</a></div><div class=\"lev2 toc-item\"><a href=\"#Prédécesseurs,-deuxième-essai\" data-toc-modified-id=\"Prédécesseurs,-deuxième-essai-113\"><span class=\"toc-item-num\">1.13&nbsp;&nbsp;</span>Prédécesseurs, deuxième essai</a></div><div class=\"lev2 toc-item\"><a href=\"#Listes\" data-toc-modified-id=\"Listes-114\"><span class=\"toc-item-num\">1.14&nbsp;&nbsp;</span>Listes</a></div><div class=\"lev2 toc-item\"><a href=\"#La-fonction-U\" data-toc-modified-id=\"La-fonction-U-115\"><span class=\"toc-item-num\">1.15&nbsp;&nbsp;</span>La fonction U</a></div><div class=\"lev2 toc-item\"><a href=\"#La-récursion-via-la-fonction-Y\" data-toc-modified-id=\"La-récursion-via-la-fonction-Y-116\"><span class=\"toc-item-num\">1.16&nbsp;&nbsp;</span>La récursion via la fonction Y</a></div><div class=\"lev2 toc-item\"><a href=\"#Conclusion\" data-toc-modified-id=\"Conclusion-117\"><span class=\"toc-item-num\">1.17&nbsp;&nbsp;</span>Conclusion</a></div>", "_____no_output_____" ], [ "# Lambda-calcul implémenté en OCaml\n\nCe notebook est inspiré de [ce post de blog du Professeur Matt Might](http://matt.might.net/articles/python-church-y-combinator/), qui implémente un mini langage de programmation en $\\lambda$-calcul, en Python.\nJe vais faire la même chose en OCaml.\n\n## Expressions\nOn rappelle que les expressions du [Lambda calcul](https://fr.wikipedia.org/wiki/Lambda-calcul), ou $\\lambda$-calcul, sont les suivantes :\n$$ \\begin{cases}\nx, y, z & \\text{(des variables)} \\\\\nu v & \\text{(application de deux termes}\\, u, v\\; \\text{)} \\\\\n\\lambda x. v & \\text{(lambda-function prenant la variable}\\; x \\;\\text{et le terme}\\; v \\;\\text{)}\n\\end{cases} $$\n\n## But ?\nLe but ne va pas être de les représenter comme ça avec des types formels en Caml, mais plutôt d'utiliser les constructions de Caml, respectivement `u(v)` et `fun x -> v` pour l'application et les fonctions anonymes, et encoder des fonctionnalités de plus haut niveau dans ce langage réduit.\n\n## Grammaire\nAvec une grammaire BNF, si `<var>` désigne un nom d'expression valide (on se limitera à des noms en miniscules consistitués des 26 lettres `a,b,..,z`) :\n\n <exp> ::= <var>\n | <exp>(<exp>)\n | fun <var> -> <exp>\n | (<exp>)", "_____no_output_____" ], [ "----\n## L'identité", "_____no_output_____" ] ], [ [ "let identite = fun x -> x ;;", "_____no_output_____" ], [ "let vide = fun x -> x ;;", "_____no_output_____" ] ], [ [ "## Conditionnelles\nLa conditionnelle est `si cond alors valeur_vraie sinon valeur_fausse`.", "_____no_output_____" ] ], [ [ "let si = fun cond valeur_vraie valeur_fausse -> cond valeur_vraie valeur_fausse ;;", "_____no_output_____" ] ], [ [ "C'est très simple, du moment qu'on s'assure que `cond` est soit `vrai` soit `faux` tels que définis par leur comportement :\n\n si vrai e1 e2 == e1\n si faux e1 e2 == e2", "_____no_output_____" ] ], [ [ "let vrai = fun valeur_vraie valeur_fausse -> valeur_vraie ;;\nlet faux = fun valeur_vraie valeur_fausse -> valeur_fausse ;;", "File \"[14]\", line 1, characters 28-41:\nWarning 27: unused variable valeur_fausse.\n" ] ], [ [ "La négation est facile !", "_____no_output_____" ] ], [ [ "let non = fun v x y -> v y x;;", "_____no_output_____" ] ], [ [ "En fait, on va forcer une évaluation paresseuse, comme ça si l'une des deux expressions ne terminent pas, l'évaluation fonctionne quand même.", "_____no_output_____" ] ], [ [ "let vrai_paresseux = fun valeur_vraie valeur_fausse -> valeur_vraie () ;;\nlet faux_paresseux = fun valeur_vraie valeur_fausse -> valeur_fausse () ;;", "File \"[16]\", line 1, characters 38-51:\nWarning 27: unused variable valeur_fausse.\n" ] ], [ [ "Pour rendre paresseux un terme, rien de plus simple !", "_____no_output_____" ] ], [ [ "let paresseux = fun f -> fun () -> f ;;", "_____no_output_____" ] ], [ [ "## Nombres\nLa représentation de Church consiste a écrire $n$ comme $\\lambda f. \\lambda z. f^n z$.", "_____no_output_____" ] ], [ [ "type 'a nombres = ('a -> 'a) -> 'a -> 'a;; (* inutilisé *)\ntype entiers_church = (int -> int) -> int -> int;;", "_____no_output_____" ] ], [ [ "$0$ est trivialement $\\lambda f. \\lambda z. z$ :", "_____no_output_____" ] ], [ [ "let zero = fun (f : ('a -> 'a)) (z : 'a) -> z ;;", "File \"[34]\", line 1, characters 16-17:\nWarning 27: unused variable f.\n" ] ], [ [ "$1$ est $\\lambda f. \\lambda z. f z$ :", "_____no_output_____" ] ], [ [ "let un = fun (f : ('a -> 'a)) -> f ;;", "_____no_output_____" ] ], [ [ "Avec l'opérateur de composition, l'écriture des entiers suivants est facile.", "_____no_output_____" ] ], [ [ "let compose = fun f g x -> f (g x);;", "_____no_output_____" ], [ "let deux = fun f -> compose f f;; (* == compose f (un f) *)\nlet trois = fun f -> compose f (deux f) ;;\nlet quatre = fun f -> compose f (trois f) ;;\n(* etc *)", "_____no_output_____" ] ], [ [ "On peut généraliser ça, avec une fonction qui transforme un entier (`int`) de Caml en un entier de Church :", "_____no_output_____" ] ], [ [ "let rec entierChurch (n : int) =\n fun f z -> if n = 0 then z else f ((entierChurch (n-1)) f z)\n;;", "_____no_output_____" ] ], [ [ "Par exemple :", "_____no_output_____" ] ], [ [ "(entierChurch 0) (fun x -> x + 1) 0;; (* 0 *)\n(entierChurch 7) (fun x -> x + 1) 0;; (* 7 *)\n(entierChurch 3) (fun x -> 2*x) 1;; (* 8 *)", "_____no_output_____" ] ], [ [ "Et une fonction qui fait l'inverse (note : cette fonction n'est *pas* un $\\lambda$-terme) :", "_____no_output_____" ] ], [ [ "let entierNatif c : int =\n c (fun x -> x + 1) 0\n;;", "_____no_output_____" ] ], [ [ "Un petit test :", "_____no_output_____" ] ], [ [ "entierNatif (si vrai zero un);; (* 0 *)\nentierNatif (si faux zero un);; (* 1 *)", "_____no_output_____" ], [ "entierNatif (entierChurch 100);; (* 100 *)", "_____no_output_____" ] ], [ [ "## Test d'inégalité\nOn a besoin de pouvoir tester si $n \\leq 0$ (ou $n = 0$) en fait.", "_____no_output_____" ] ], [ [ "(* prend un lambda f lambda z. ... est donne vrai ssi n = 0 ou faux sinon *)\nlet estnul = fun n -> n (fun z -> faux) (vrai);;", "File \"[43]\", line 2, characters 29-30:\nWarning 27: unused variable z.\n" ], [ "(* prend un lambda f lambda z. ... est donne vrai ssi n > 0 ou faux sinon *)\nlet estnonnul = fun n -> n (fun z -> vrai) (faux);;", "File \"[44]\", line 2, characters 32-33:\nWarning 27: unused variable z.\n" ] ], [ [ "On peut proposer cette autre implémentation, qui \"fonctionne\" pareil (au sens calcul des $\\beta$-réductions) mais est plus compliquée :", "_____no_output_____" ] ], [ [ "let estnonnul2 = fun n -> non (estnul n);;", "_____no_output_____" ], [ "entierNatif (si (estnul zero) zero un);; (* 0 *)\nentierNatif (si (estnul un) zero un);; (* 1 *)\nentierNatif (si (estnul deux) zero un);; (* 1 *)", "_____no_output_____" ], [ "entierNatif (si (estnonnul zero) zero un);; (* 0 *)\nentierNatif (si (estnonnul un) zero un);; (* 1 *)\nentierNatif (si (estnonnul deux) zero un);; (* 1 *)", "_____no_output_____" ], [ "entierNatif (si (non (estnul zero)) zero un);; (* 0 *)\nentierNatif (si (non (estnul un)) zero un);; (* 1 *)\nentierNatif (si (non (estnul deux)) zero un);; (* 1 *)", "_____no_output_____" ] ], [ [ "## Successeurs\nVue la représentation de Churc, $n+1$ consiste a appliquer l'argument $f$ une fois de plus :\n$f^{n+1}(z) = f (f^n(z))$.", "_____no_output_____" ] ], [ [ "let succ = fun n f z -> f ((n f) z) ;;", "_____no_output_____" ], [ "entierNatif (succ un);; (* 2 *)", "_____no_output_____" ], [ "deux;;\nsucc un;;", "_____no_output_____" ] ], [ [ "On remarque qu'ils ont le même typage, mais OCaml indique qu'il a moins d'informations à propos du deuxième : ce `'_a` signifie que le type est *contraint*, il sera fixé dès la première utilisation de cette fonction.\n\nC'est assez mystérieux, mais il faut retenir le point suivant : `deux` était écrit manuellement, donc le système a vu le terme en entier, il le connaît et saît que `deux = fun f -> fun x -> f (f x))`, pas de surprise. Par contre, `succ un` est le résultat d'une évaluation *partielle* et vaut `fun f z -> f ((deux f) z)`. Sauf que le système ne calcule pas tout et laisse l'évaluation partielle ! (heureusement !)", "_____no_output_____" ], [ "Si on appelle `succ un` à une fonction, le `'_a` va être contraint, et on ne pourra pas s'en reservir :", "_____no_output_____" ] ], [ [ "let succ_de_un = succ un;;", "_____no_output_____" ], [ "(succ_de_un) (fun x -> x + 1);;", "_____no_output_____" ], [ "(succ_de_un) (fun x -> x ^ \"0\");;", "_____no_output_____" ], [ "(succ un) (fun x -> x ^ \"0\");;\n(* une valeur fraîchement calculée, sans contrainte *)", "_____no_output_____" ] ], [ [ "## Prédecesseurs\nVue la représentation de Church, $\\lambda n. n-1$ n'existe pas... mais on peut tricher.", "_____no_output_____" ] ], [ [ "let pred = fun n ->\n if (entierNatif n) > 0 then entierChurch ((entierNatif n) - 1)\n else zero\n;;", "_____no_output_____" ], [ "entierNatif (pred deux);; (* 1 *)", "_____no_output_____" ], [ "entierNatif (pred trois);; (* 2 *)", "_____no_output_____" ] ], [ [ "## Addition\n\nPour ajouter $n$ et $m$, il faut appliquer une fonction $f$ $n$ fois puis $m$ fois : $f^{n+m}(z) = f^n(f^m(z))$.", "_____no_output_____" ] ], [ [ "let somme = fun n m f z -> n(f)( m(f)(z));;", "_____no_output_____" ], [ "let cinq = somme deux trois ;;", "_____no_output_____" ], [ "entierNatif cinq;;", "_____no_output_____" ], [ "let sept = somme cinq deux ;;", "_____no_output_____" ], [ "entierNatif sept;;", "_____no_output_____" ] ], [ [ "## Multiplication\n\nPour multiplier $n$ et $m$, il faut appliquer le codage de $n$ exactement $m$ fois : $f^{nm}(z) = (f^n(f^n(...(f^n(z))...))$.", "_____no_output_____" ] ], [ [ "let produit = fun n m f z -> m(n(f))(z);;", "_____no_output_____" ] ], [ [ "On peut faire encore mieux avec l'opérateur de composition :", "_____no_output_____" ] ], [ [ "let produit = fun n m -> compose m n;;", "_____no_output_____" ], [ "let six = produit deux trois ;;", "_____no_output_____" ], [ "entierNatif six;;", "_____no_output_____" ], [ "let huit = produit deux quatre ;;", "_____no_output_____" ], [ "entierNatif huit;;", "_____no_output_____" ] ], [ [ "## Paires\n\nOn va écrire un constructeur de paires, `paire a b` qui sera comme `(a, b)`, et deux destructeurs, `gauche` et `droite`, qui vérifient :\n\n gauche (paire a b) == a\n droite (paire a b) == b", "_____no_output_____" ] ], [ [ "let paire = fun a b -> fun f -> f(a)(b);;", "_____no_output_____" ], [ "let gauche = fun p -> p(fun a b -> a);;\nlet droite = fun p -> p(fun a b -> b);;", "File \"[76]\", line 1, characters 30-31:\nWarning 27: unused variable b.\n" ], [ "entierNatif (gauche (paire zero un));;\nentierNatif (droite (paire zero un));;", "_____no_output_____" ] ], [ [ "## Prédécesseurs, deuxième essai\n\nIl y a une façon, longue et compliquée ([source](http://gregfjohnson.com/pred/)) d'y arriver, avec des paires.", "_____no_output_____" ] ], [ [ "let pred n suivant premier =\n let pred_suivant = paire vrai premier in\n let pred_premier = fun p ->\n si (gauche p)\n (paire faux premier)\n (paire faux (suivant (droite p)))\n in\n let paire_finale = n pred_suivant pred_premier in\n droite paire_finale\n;;", "_____no_output_____" ] ], [ [ "Malheureusement, ce n'est pas bien typé.", "_____no_output_____" ] ], [ [ "entierNatif (pred deux);; (* 1 *)", "_____no_output_____" ] ], [ [ "## Listes\n\nPour construire des listes (simplement chaînées), on a besoin d'une valeur pour la liste vide, `listevide`, d'un constructeur pour une liste `cons`, un prédicat pour la liste vide `estvide`, un accesseur `tete` et `queue`, et avec les contraintes suivantes (avec `vrai`, `faux` définis comme plus haut) :\n\n estvide (listevide) == vrai\n estvide (cons tt qu) == faux\n \n tete (cons tt qu) == tt\n queue (cons tt qu) == qu\n\nOn va stocker tout ça avec des fonctions qui attendront deux arguments (deux fonctions - rappel tout est fonction en $\\lambda$-calcul), l'une appellée si la liste est vide, l'autre si la liste n'est pas vide.", "_____no_output_____" ] ], [ [ "let listevide = fun survide surpasvide -> survide;;", "File \"[58]\", line 1, characters 28-38:\nWarning 27: unused variable surpasvide.\n" ], [ "let cons = fun hd tl -> fun survide surpasvide -> surpasvide hd tl;;", "_____no_output_____" ] ], [ [ "Avec cette construction, `estvide` est assez simple : `survide` est `() -> vrai` et `surpasvide` est `tt qu -> faux`.", "_____no_output_____" ] ], [ [ "let estvide = fun liste -> liste (vrai) (fun tt qu -> faux);;", "File \"[60]\", line 1, characters 45-47:\nWarning 27: unused variable tt.\nFile \"[60]\", line 1, characters 48-50:\nWarning 27: unused variable qu.\n" ] ], [ [ "Deux tests :", "_____no_output_____" ] ], [ [ "entierNatif (si (estvide (listevide)) un zero);; (* estvide listevide == vrai *)\nentierNatif (si (estvide (cons un listevide)) un zero);; (* estvide (cons un listevide) == faux *)", "_____no_output_____" ] ], [ [ "Et pour les deux extracteurs, c'est très facile avec cet encodage.", "_____no_output_____" ] ], [ [ "let tete = fun liste -> liste (vide) (fun tt qu -> tt);;\nlet queue = fun liste -> liste (vide) (fun tt qu -> qu);;", "File \"[62]\", line 1, characters 45-47:\nWarning 27: unused variable qu.\n" ], [ "entierNatif (tete (cons un listevide));;\nentierNatif (tete (queue (cons deux (cons un listevide))));;\nentierNatif (tete (queue (cons trois (cons deux (cons un listevide)))));;", "_____no_output_____" ] ], [ [ "Visualisons les types que Caml trouve a des listes de tailles croissantes :", "_____no_output_____" ] ], [ [ "cons un (cons un listevide);; (* 8 variables pour une liste de taille 2 *)", "_____no_output_____" ], [ "cons un (cons un (cons un (cons un listevide)));; (* 14 variables pour une liste de taille 4 *)", "_____no_output_____" ], [ "cons un (cons un (cons un (cons un (cons un (cons un (cons un (cons un listevide)))))));; (* 26 variables pour une liste de taille 7 *)", "_____no_output_____" ] ], [ [ "Pour ces raisons là, on se rend compte que le type donné par Caml à une liste de taille $k$ croît linéairement *en taille* en fonction de $k$ !\n\nAucun espoir donc (avec cet encodage) d'avoir un type générique pour les listes représentés en Caml.\n\nEt donc nous ne sommes pas surpris de voir cet essai échouer :", "_____no_output_____" ] ], [ [ "let rec longueur liste =\n liste (zero) (fun t q -> succ (longueur q))\n;;", "_____no_output_____" ] ], [ [ "<span style=\"color:red;\">En effet, `longueur` devrait être bien typée et `liste` et `q` devraient avoir le même type, or le type de `liste` est strictement plus grand que celui de `q`...</span>", "_____no_output_____" ], [ "On peut essayer de faire une fonction `ieme`.\nOn veut que `ieme zero liste = tete` et `ieme n liste = ieme (pred n) (queue liste)`.\n\nEn écrivant en haut niveau, on aimerait pouvoir faire :", "_____no_output_____" ] ], [ [ "let pop liste =\n si (estvide liste) (listevide) (queue liste)\n;;", "_____no_output_____" ], [ "let ieme n liste =\n tete (n pop liste)\n;;", "_____no_output_____" ] ], [ [ "## La fonction U\n\nC'est le premier indice que le $\\lambda$-calcul peut être utilisé comme modèle de calcul : le terme $U : f \\to f(f)$ ne termine pas si on l'applique à lui-même.\n\nMais ce sera la faiblesse de l'utilisation de Caml : ce terme ne peut être correctement typé !", "_____no_output_____" ] ], [ [ "let u = fun f -> f (f);;", "_____no_output_____" ] ], [ [ "A noter que même dans un langage non typé (par exemple Python), on peut définir $U$ mais son exécution échouera, soit à caude d'un dépassement de pile, soit parce qu'elle ne termine pas.", "_____no_output_____" ], [ "## La récursion via la fonction Y\n\nLa fonction $Y$ trouve le point fixe d'une autre fonction.\nC'est très utile pour définir des fonctions par récurrence.\n\nPar exemple, la factorielle est le point fixe de la fonction suivante :\n\"$\\lambda f. \\lambda n. 1$ si $n \\leq 0$ sinon $n * f(n-1)$\" (écrite dans un langage plus haut niveau, pas en $\\lambda$-calcul).\n\n$Y$ satisfait ces contraintes : $Y(F) = f$ et $f = F(f)$.\nDonc $Y(F) = F(Y(F))$ et donc $Y = \\lambda F. F(Y(F))$. Mais ce premier essai ne marche pas.", "_____no_output_____" ] ], [ [ "let rec y = fun f -> f (y(f));;", "_____no_output_____" ], [ "let fact = y(fun f n -> si (estnul n) (un) (produit n (f (pred n))));;", "_____no_output_____" ] ], [ [ "On utilise la $\\eta$-expansion : si $e$ termine, $e$ est équivalent (ie tout calcul donne le même terme) à $\\lambda x. e(x)$.", "_____no_output_____" ] ], [ [ "let rec y = fun f -> f (fun x -> y(f)(x));;", "_____no_output_____" ] ], [ [ "Par contre, le typage n'arrive toujours pas à trouver que l'expression suivante devrait être bien définie :", "_____no_output_____" ] ], [ [ "let fact = y(fun f n -> si (estnul n) (un) (produit n (f (pred n))));;", "_____no_output_____" ] ], [ [ "## Conclusion\n\nJe n'ai pas réussi à traduire intégralement la prouesse initiale, écrite en Python, par Matt Might.\nDommage, le typage de Caml est trop strict pour cet exercice.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# %matplotlib inline", "_____no_output_____" ], [ "import seaborn as sns\nimport pandas as pd\nimport numpy as np\nimport matplotlib as mpl\nimport matplotlib.pyplot as plt", "_____no_output_____" ], [ "list_2d = [[0, 1, 2], [3, 4, 5]]", "_____no_output_____" ], [ "plt.figure()\nsns.heatmap(list_2d)\nplt.savefig('data/dst/seaborn_heatmap_list.png')\nplt.close('all')", "_____no_output_____" ], [ "arr_2d = np.arange(-8, 8).reshape((4, 4))\nprint(arr_2d)", "[[-8 -7 -6 -5]\n [-4 -3 -2 -1]\n [ 0 1 2 3]\n [ 4 5 6 7]]\n" ], [ "plt.figure()\nsns.heatmap(arr_2d)\nplt.savefig('data/dst/seaborn_heatmap_ndarray.png')", "_____no_output_____" ], [ "df = pd.DataFrame(data=arr_2d, index=['a', 'b', 'c', 'd'], columns=['A', 'B', 'C', 'D'])\nprint(df)", " A B C D\na -8 -7 -6 -5\nb -4 -3 -2 -1\nc 0 1 2 3\nd 4 5 6 7\n" ], [ "plt.figure()\nsns.heatmap(df)\nplt.savefig('data/dst/seaborn_heatmap_dataframe.png')", "_____no_output_____" ], [ "print(type(sns.heatmap(list_2d)))", "<class 'matplotlib.axes._subplots.AxesSubplot'>\n" ], [ "fig = plt.figure()\nax = fig.add_subplot(1, 1, 1)\nsns.heatmap(list_2d, ax=ax)\nfig.savefig('data/dst/seaborn_heatmap_list.png')", "_____no_output_____" ], [ "fig, axes = plt.subplots(nrows=2, ncols=3, figsize=(8, 6))\nsns.heatmap(list_2d, ax=axes[0, 0])\nsns.heatmap(arr_2d, ax=axes[1, 2])\nfig.savefig('data/dst/seaborn_heatmap_list_sub.png')", "_____no_output_____" ], [ "plt.figure()\nsns.heatmap(df, annot=True)\nplt.savefig('data/dst/seaborn_heatmap_annot.png')", "_____no_output_____" ], [ "plt.figure()\nsns.heatmap(df, cbar=False)\nplt.savefig('data/dst/seaborn_heatmap_no_cbar.png')", "_____no_output_____" ], [ "plt.figure()\nsns.heatmap(df, square=True)\nplt.savefig('data/dst/seaborn_heatmap_square.png')", "_____no_output_____" ], [ "plt.figure()\nsns.heatmap(df, vmax=10, vmin=-10, center=0)\nplt.savefig('data/dst/seaborn_heatmap_vmax_vmin_center.png')", "_____no_output_____" ], [ "plt.figure()\nsns.heatmap(df, cmap='hot')\nplt.savefig('data/dst/seaborn_heatmap_hot.png')", "_____no_output_____" ], [ "plt.figure()\nsns.heatmap(df, cmap='Blues')\nplt.savefig('data/dst/seaborn_heatmap_blues.png')", "_____no_output_____" ], [ "plt.figure()\nsns.heatmap(df, cmap='Blues_r')\nplt.savefig('data/dst/seaborn_heatmap_blues_r.png')", "_____no_output_____" ], [ "current_figsize = mpl.rcParams['figure.figsize']\nprint(current_figsize)", "[6.0, 4.0]\n" ], [ "plt.figure(figsize=(9, 6)) \nsns.heatmap(df, square=True)\nplt.savefig('data/dst/seaborn_heatmap_big.png')", "_____no_output_____" ], [ "current_dpi = mpl.rcParams['figure.dpi']\nprint(current_dpi)", "72.0\n" ], [ "plt.figure()\nsns.heatmap(df, square=True)\nplt.savefig('data/dst/seaborn_heatmap_big_2.png', dpi=current_dpi * 1.5)", "_____no_output_____" ], [ "df_house = pd.read_csv('data/src/house_prices_train.csv', index_col=0)", "_____no_output_____" ], [ "df_house_corr = df_house.corr()\nprint(df_house_corr.shape)", "(37, 37)\n" ], [ "print(df_house_corr.head())", " MSSubClass LotFrontage LotArea OverallQual OverallCond \\\nMSSubClass 1.000000 -0.386347 -0.139781 0.032628 -0.059316 \nLotFrontage -0.386347 1.000000 0.426095 0.251646 -0.059213 \nLotArea -0.139781 0.426095 1.000000 0.105806 -0.005636 \nOverallQual 0.032628 0.251646 0.105806 1.000000 -0.091932 \nOverallCond -0.059316 -0.059213 -0.005636 -0.091932 1.000000 \n\n YearBuilt YearRemodAdd MasVnrArea BsmtFinSF1 BsmtFinSF2 \\\nMSSubClass 0.027850 0.040581 0.022936 -0.069836 -0.065649 \nLotFrontage 0.123349 0.088866 0.193458 0.233633 0.049900 \nLotArea 0.014228 0.013788 0.104160 0.214103 0.111170 \nOverallQual 0.572323 0.550684 0.411876 0.239666 -0.059119 \nOverallCond -0.375983 0.073741 -0.128101 -0.046231 0.040229 \n\n ... WoodDeckSF OpenPorchSF EnclosedPorch 3SsnPorch \\\nMSSubClass ... -0.012579 -0.006100 -0.012037 -0.043825 \nLotFrontage ... 0.088521 0.151972 0.010700 0.070029 \nLotArea ... 0.171698 0.084774 -0.018340 0.020423 \nOverallQual ... 0.238923 0.308819 -0.113937 0.030371 \nOverallCond ... -0.003334 -0.032589 0.070356 0.025504 \n\n ScreenPorch PoolArea MiscVal MoSold YrSold SalePrice \nMSSubClass -0.026030 0.008283 -0.007683 -0.013585 -0.021407 -0.084284 \nLotFrontage 0.041383 0.206167 0.003368 0.011200 0.007450 0.351799 \nLotArea 0.043160 0.077672 0.038068 0.001205 -0.014261 0.263843 \nOverallQual 0.064886 0.065166 -0.031406 0.070815 -0.027347 0.790982 \nOverallCond 0.054811 -0.001985 0.068777 -0.003511 0.043950 -0.077856 \n\n[5 rows x 37 columns]\n" ], [ "plt.figure(figsize=(12, 9)) \nsns.heatmap(df_house_corr, square=True, vmax=1, vmin=-1, center=0)\nplt.savefig('data/dst/seaborn_heatmap_house_price.png')", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<h1>Table of Contents<span class=\"tocSkip\"></span></h1>\n<div class=\"toc\"><ul class=\"toc-item\"><li><span><a href=\"#Dimensionality-Reduction\" data-toc-modified-id=\"Dimensionality-Reduction-1\"><span class=\"toc-item-num\">1&nbsp;&nbsp;</span>Dimensionality Reduction</a></span><ul class=\"toc-item\"><li><span><a href=\"#The-Problem\" data-toc-modified-id=\"The-Problem-1.1\"><span class=\"toc-item-num\">1.1&nbsp;&nbsp;</span>The Problem</a></span><ul class=\"toc-item\"><li><span><a href=\"#Multi-Collinearity\" data-toc-modified-id=\"Multi-Collinearity-1.1.1\"><span class=\"toc-item-num\">1.1.1&nbsp;&nbsp;</span>Multi-Collinearity</a></span></li></ul></li><li><span><a href=\"#Sparsity\" data-toc-modified-id=\"Sparsity-1.2\"><span class=\"toc-item-num\">1.2&nbsp;&nbsp;</span>Sparsity</a></span></li></ul></li><li><span><a href=\"#Principle-Component-Analysis\" data-toc-modified-id=\"Principle-Component-Analysis-2\"><span class=\"toc-item-num\">2&nbsp;&nbsp;</span>Principle Component Analysis</a></span><ul class=\"toc-item\"><li><span><a href=\"#Important-Points:\" data-toc-modified-id=\"Important-Points:-2.1\"><span class=\"toc-item-num\">2.1&nbsp;&nbsp;</span>Important Points:</a></span></li></ul></li><li><span><a href=\"#Singular-Value-Decomposition\" data-toc-modified-id=\"Singular-Value-Decomposition-3\"><span class=\"toc-item-num\">3&nbsp;&nbsp;</span>Singular Value Decomposition</a></span><ul class=\"toc-item\"><li><span><a href=\"#Measuring-the-Quality-of-the-Reconstruction\" data-toc-modified-id=\"Measuring-the-Quality-of-the-Reconstruction-3.1\"><span class=\"toc-item-num\">3.1&nbsp;&nbsp;</span>Measuring the Quality of the Reconstruction</a></span></li><li><span><a href=\"#Heuristic-Step-for-How-Many-Dimensions-to-Keep\" data-toc-modified-id=\"Heuristic-Step-for-How-Many-Dimensions-to-Keep-3.2\"><span class=\"toc-item-num\">3.2&nbsp;&nbsp;</span>Heuristic Step for How Many Dimensions to Keep</a></span></li></ul></li><li><span><a href=\"#GLOVE\" data-toc-modified-id=\"GLOVE-4\"><span class=\"toc-item-num\">4&nbsp;&nbsp;</span>GLOVE</a></span><ul class=\"toc-item\"><li><span><a href=\"#Using-Spacy-word2vec-embeddings\" data-toc-modified-id=\"Using-Spacy-word2vec-embeddings-4.1\"><span class=\"toc-item-num\">4.1&nbsp;&nbsp;</span>Using Spacy word2vec embeddings</a></span></li><li><span><a href=\"#Using-Glove\" data-toc-modified-id=\"Using-Glove-4.2\"><span class=\"toc-item-num\">4.2&nbsp;&nbsp;</span>Using Glove</a></span></li></ul></li><li><span><a href=\"#Clustering-Text\" data-toc-modified-id=\"Clustering-Text-5\"><span class=\"toc-item-num\">5&nbsp;&nbsp;</span>Clustering Text</a></span></li></ul></div>", "_____no_output_____" ], [ "# Dimensionality Reduction\n\n## The Problem\n\nThere is an interesting tradeoff between model performance and a feature's dimensionality:\n\n\n>*If the amount of available training data is fixed, then overfitting occurs if we keep adding dimensions. On the other hand, if we keep adding dimensions, the amount of **training data needs to grow exponentially fast to maintain the same coverage** and to avoid overfitting* ([Computer Vision for Dummies](http://www.visiondummy.com/2014/04/curse-dimensionality-affect-classification/)).\n\n\n\n### Multi-Collinearity\n\nIn many cases, there is a high degree of correlation between many of the features in a dataset. This multi-collinearity has the effect of drowning out the \"signal\" of your dataset in many cases, and amplifies \"outlier\" noise.\n\n\n## Sparsity\n\n- High dimensionality increases the sparsity of your features (**what NLP techniques have we used that illustrate this point?**)\n- The density of the training samples decreases when dimensionality increases:\n- **Distance measures (Euclidean, for instance) start losing their effectiveness**, because there isn't much difference between the max and min distances in higher dimensions.\n- Many models that rely upon **assumptions of Gaussian distributions** (like OLS linear regression), Gaussian mixture models, Gaussian processes, etc. become less and less effective since their distributions become flatter and \"fatter tailed\".\n\n\nWhat is the amount of data needed to maintain **20% coverage** of the feature space? For 1 dimension, it is **20% of the entire population's dataset**. For a dimensionality of $D$:\n\n$$\nX^{D} = .20\n$$\n$$\n(X^{D})^{\\frac{1}{D}} = .20^{\\frac{1}{D}}\n$$\n$$\nX = \\sqrt[D]{.20}\n$$\nYou can approximate this as \n```python\ndef coverage_requirement(requirement, D):\n return requirement ** (1 / D)\n\nx = []\ny = []\nfor d in range(1,20):\n y.append(coverage_requirement(0.10, d))\n x.append(d)\n \nimport matplotlib.pyplot as plt\n\nplt.plot(x,y)\nplt.xlabel(\"Number of Dimensions\")\nplt.ylabel(\"Appromximate % of Population Dataset\")\nplt.title(\"% of Dataset Needed to Maintain 10% Coverage of Feature Space\")\nplt.show()\n```\n<img src=\"images/coverage-needed.png\" width=\"500\">", "_____no_output_____" ] ], [ [ "import pandas as pd\nfrom sklearn.feature_extraction.text import CountVectorizer\nreviews = pd.read_csv(\"mcdonalds-yelp-negative-reviews.csv\", encoding='latin-1')\nreviews = open(\"poor_amazon_toy_reviews.txt\", encoding='latin-1')\n\n#text = reviews[\"review\"].values\ntext = reviews.readlines()\n\nvectorizer = CountVectorizer(ngram_range=(3,3), min_df=0.01, max_df=0.75, max_features=200)\n# tokenize and build vocab\nvectorizer.fit(text)\nvector = vectorizer.transform(text)\nfeatures = vector.toarray()\nfeatures_df = pd.DataFrame(features, columns=vectorizer.get_feature_names())\n\ncorrelations = features_df.corr()\ncorrelations_stacked = correlations.stack().reset_index()\n#set column names\ncorrelations_stacked.columns = ['Tri-Gram 1','Tri-Gram 2','Correlation']\ncorrelations_stacked = correlations_stacked[correlations_stacked[\"Correlation\"] < 1]\ncorrelations_stacked = correlations_stacked.sort_values(by=['Correlation'], ascending=False)\ncorrelations_stacked.head()", "_____no_output_____" ], [ "import numpy as np\nimport matplotlib.pyplot as plt\n# visualize the correlations (install seaborn first)!\nimport seaborn as sns\n\n# Generate a mask for the upper triangle\nmask = np.triu(np.ones_like(correlations, dtype=np.bool))\n\n# Set up the matplotlib figure\nf, ax = plt.subplots(figsize=(11, 9))\n\n# Generate a custom diverging colormap\ncmap = sns.diverging_palette(220, 10, as_cmap=True)\n\n# Draw the heatmap with the mask and correct aspect ratio\nsns.heatmap(correlations, mask=mask, cmap=cmap, vmax=.3, center=0,\n square=True, linewidths=.5, cbar_kws={\"shrink\": .5})\n\n", "_____no_output_____" ] ], [ [ "# Principle Component Analysis\n\nIf you have an original matrix $Z$, you can decompose this matrix into two smaller matrices $X$ and $Q$. \n\n## Important Points:\n\n- Multiplying a vector by a matrix typically changes the direction of the vector. For instance:\n<figure>\n <img src=\"images/multvector.png\" alt=\"my alt text\"/>\n <figcaption><a href=\"https://lazyprogrammer.me/tutorial-principal-components-analysis-pca\">Lazy Programmer- \n Tutorial to PCA</a></figcaption>\n</figure>", "_____no_output_____" ], [ "However, there are eigenvalues λ and eigenvectors $v$ such that\n\n$$\n\\sum_{X}v = \\lambda v\n$$\n\nMultiplying the eigenvectors $v$ with the eigenvalue $\\lambda$ does not change the direction of the eigenvector.\n\nMultiplying the eigenvector $v$ by the covariance matrix $\\sum_{X}$ also does not change the direction of the eigenvector.\n\nIf our data $X$ is of shape $N \\times D$, it turns out that we have $D$ eigenvalues and $D$ eigenvectors. This means we can arrange the eigenvalues $\\lambda$ in decreasing order so that\n\n$$\n\\lambda_3 > \\lambda_2 > \\lambda_5\n$$\n\nIn this case, $\\lambda_3$ is the largest eigenvalue, followed by $\\lambda_2$, and then $\\lambda_5$. Then, we can arrange \n\nWe can also rearrange the eigenvectors the same: $v_3$ will be the first column, $v_2$ will be the second column, and $v_5$ will be the third column.\n\nWe'll end up with two matrices $V$ and $\\Lambda$:\n<figure>\n <img src=\"images/pca1.png\" alt=\"my alt text\"/>\n <figcaption><a href=\"https://lazyprogrammer.me/tutorial-principal-components-analysis-pca\">Lazy Programmer- \n Tutorial to PCA</a></figcaption>\n</figure>", "_____no_output_____" ] ], [ [ "# what is the shape of our features?\nfeatures.shape", "_____no_output_____" ], [ "from sklearn.decomposition import PCA\npca = PCA(n_components=4)\nZ = pca.fit_transform(features)\n\n# what is the shape of Z?\nZ.shape", "_____no_output_____" ], [ "# what will happen if we take the correlation matrix and covariance matrix of our new reduced features?\nimport numpy as np\ncovariances = pd.DataFrame(np.cov(Z.transpose()))\nplt.rcParams[\"figure.figsize\"] = (5,5)\nsns.heatmap(covariances)", "_____no_output_____" ], [ "# train the model to reduce the dimensions down to 2\npca = PCA(n_components=2)\nZ_two_dimensions = pca.fit_transform(features)\nZ_two_dimensions", "_____no_output_____" ], [ "import matplotlib.pyplot as plt\n\nplt.scatter(Z_two_dimensions[:,0], Z_two_dimensions[:, 1])\nreduced_features_df = pd.DataFrame(Z_two_dimensions, columns=[\"x1\", \"x2\"])\nreduced_features_df[\"text\"] = text", "_____no_output_____" ] ], [ [ "# Singular Value Decomposition\n\nGiven an input matrix $A$, we want to try to represent it instead as three smaller matrices $U$, $\\sum$, and $V$. Instead of **$n$ original terms**, we want to represent each document as **$r$ concepts** (other referred to as **latent dimensions**, or **latent factors**):\n<figure>\n <img src=\"images/svd.png\" alt=\"my alt text\"/>\n <figcaption><i>\n <a href=\"https://www.youtube.com/watch?v=P5mlg91as1c\">Mining of Massive Datasets - Dimensionality Reduction: Singular Value Decomposition</a> by Leskovec, Rajaraman, and Ullman (Stanford University)</i></figcaption>\n</figure>\n\nHere, **$A$ is your matrix of word vectors** - you could use any of the word vectorization techniques we have learned so far, include one-hot encoding, word count, TF-IDF.\n\n- $\\sum$ will be a **diagonal matrix** with values that are positive and sorted in decreasing order. Its value indicate the **variance (information encoded on that new dimension)**- therefore, the higher the value, the stronger that dimension is in capturing data from A, the original features. For our purposes, we can think of the rank of this $\\sum$ matrix as the number of desired dimensions. Instance, if we want to reduce $A$ from shape $1020 x 300$ to $1020 x 10$, we will want to reduce the rank of $\\sum$ from 300 to 10.\n\n- $U^T U = I$ and $V^T V = I$\n\n## Measuring the Quality of the Reconstruction\n\nA popular metric used for measuring the quality of the reconstruction is the [Frobenius Norm](https://en.wikipedia.org/wiki/Matrix_norm#Frobenius_norm). When you explain your methodology for reducing dimensions, usually managers / stakeholders will want to some way to compare multiple dimensionality techniques' ability to quantify its ability to retain information but trim dimensions:\n\n$$\n\\begin{equation}\n||A_{old}-A_{new}||_{F} = \\sqrt{\\sum_{ij}{(A^{old}_{ij}- A^{new}_{ij}}})^2\n\\end{equation}\n$$\n\n## Heuristic Step for How Many Dimensions to Keep\n\n1. Sum the $\\sum$ matrix's diagonal values: \n$$\n\\begin{equation}\n\\sum_{i}^{m}\\sigma_{i}\n\\end{equation}\n$$\n\n2. Define your threshold of \"information\" (variance) $\\alpha$ to keep: usually 80% to 90%. \n\n3. Define your cutoff point $C$: $$\n\\begin{equation}\nC = \\sum_{i}^{m}\\sigma_{i} \\alpha\n\\end{equation}\n$$\n\n4. Beginning with your largest singular value, sum your singular values $\\sigma_{i}$ until it is greater than C. Retain only those dimensions.", "_____no_output_____" ], [ "<figure>\n <img src=\"images/userratings.png\" alt=\"my alt text\"/>\n <figcaption><i>\n <a href=\"https://www.youtube.com/watch?v=P5mlg91as1c\">Mining of Massive Datasets - Dimensionality Reduction: Singular Value Decomposition</a> by Leskovec, Rajaraman, and Ullman (Stanford University)</i></figcaption>\n</figure>\n", "_____no_output_____" ] ], [ [ "# create sample data\n\nimport numpy as np\nimport matplotlib.pyplot as plt\nfrom scipy.linalg import svd\nx = np.linspace(1,20, 20) # create the first dimension\nx = np.concatenate((x,x))\ny = x + np.random.normal(0,1, 40) # create the second dimension\nz = x + np.random.normal(0,2, 40) # create the third dimension\na = x + np.random.normal(0,4, 40) # create the fourth dimension\nplt.scatter(x,y) # plot just the first two dimensions\nplt.show()", "_____no_output_____" ], [ "# create matrix\nA = np.stack([x,y,z,a]).T", "_____no_output_____" ], [ "# perform SVD\nD = 1\nU, s, V = svd(A)\nprint(f\"s is {s}\\n\")\nprint(f\"U is {U}\\n\")\nprint(f\"V is {V}\")", "_____no_output_____" ], [ "# Frobenius norm\n\ns[D:] = 0\nS = np.zeros((A.shape[0], A.shape[1]))\nS[:A.shape[1], :A.shape[1]] = np.diag(s)\nA_reconstructed = U.dot(S.dot(V))\nnp.sum((A_reconstructed - A) ** 2) ** (1/2) # Frobenius norm\n# reconstruct matrix\nU.dot(S)", "_____no_output_____" ] ], [ [ "# GLOVE\n\nGlobal vectors for word presentation:\n<figure>\n <img src=\"images/glove_1.png\" alt=\"my alt text\"/>\n <figcaption><i>\n <a href=\"https://nlp.stanford.edu/pubs/glove.pdf\">GloVe: Global Vectors for Word Representation</a></i></figcaption>\n</figure>", "_____no_output_____" ] ], [ [ "!pip3 install gensim", "Collecting gensim\n Downloading gensim-3.8.2.tar.gz (23.4 MB)\n\u001b[K |████████████████████████████████| 23.4 MB 187 kB/s eta 0:00:01\n\u001b[?25hRequirement already satisfied: numpy>=1.11.3 in /usr/local/lib/python3.8/site-packages (from gensim) (1.18.2)\nRequirement already satisfied: scipy>=1.0.0 in /usr/local/lib/python3.8/site-packages (from gensim) (1.4.1)\nRequirement already satisfied: six>=1.5.0 in /usr/local/lib/python3.8/site-packages (from gensim) (1.14.0)\nCollecting smart_open>=1.8.1\n Downloading smart_open-1.11.1.tar.gz (105 kB)\n\u001b[K |████████████████████████████████| 105 kB 264 kB/s eta 0:00:01\n\u001b[?25hRequirement already satisfied: requests in /usr/local/lib/python3.8/site-packages (from smart_open>=1.8.1->gensim) (2.23.0)\nCollecting boto\n Downloading boto-2.49.0-py2.py3-none-any.whl (1.4 MB)\n\u001b[K |████████████████████████████████| 1.4 MB 205 kB/s eta 0:00:01\n\u001b[?25hRequirement already satisfied: boto3 in /usr/local/lib/python3.8/site-packages (from smart_open>=1.8.1->gensim) (1.12.39)\nRequirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.8/site-packages (from requests->smart_open>=1.8.1->gensim) (2020.4.5.1)\nRequirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.8/site-packages (from requests->smart_open>=1.8.1->gensim) (2.9)\nRequirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.8/site-packages (from requests->smart_open>=1.8.1->gensim) (3.0.4)\nRequirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.8/site-packages (from requests->smart_open>=1.8.1->gensim) (1.25.8)\nRequirement already satisfied: botocore<1.16.0,>=1.15.39 in /usr/local/lib/python3.8/site-packages (from boto3->smart_open>=1.8.1->gensim) (1.15.39)\nRequirement already satisfied: s3transfer<0.4.0,>=0.3.0 in /usr/local/lib/python3.8/site-packages (from boto3->smart_open>=1.8.1->gensim) (0.3.3)\nRequirement already satisfied: jmespath<1.0.0,>=0.7.1 in /usr/local/lib/python3.8/site-packages (from boto3->smart_open>=1.8.1->gensim) (0.9.5)\nRequirement already satisfied: python-dateutil<3.0.0,>=2.1 in /usr/local/lib/python3.8/site-packages (from botocore<1.16.0,>=1.15.39->boto3->smart_open>=1.8.1->gensim) (2.8.1)\nRequirement already satisfied: docutils<0.16,>=0.10 in /usr/local/lib/python3.8/site-packages (from botocore<1.16.0,>=1.15.39->boto3->smart_open>=1.8.1->gensim) (0.15.2)\nBuilding wheels for collected packages: gensim, smart-open\n Building wheel for gensim (setup.py) ... \u001b[?25ldone\n\u001b[?25h Created wheel for gensim: filename=gensim-3.8.2-cp38-cp38-macosx_10_14_x86_64.whl size=24208273 sha256=6bd62ae0639fb3236fa9f41540baf674913434b2bcb4124b691f9bdd7a76b93e\n Stored in directory: /Users/yuchen/Library/Caches/pip/wheels/91/fa/af/b14a7e3943121f028c82af23946002f99ba18853de2381b148\n Building wheel for smart-open (setup.py) ... \u001b[?25ldone\n\u001b[?25h Created wheel for smart-open: filename=smart_open-1.11.1-py3-none-any.whl size=95255 sha256=a104018d08730ff5b1f115c5e2cbf0c58a0becb891a33e3e3bb9f4060fac5705\n Stored in directory: /Users/yuchen/Library/Caches/pip/wheels/90/f7/ab/7b855402a7d4bf9dececf03bb7f8c831dabad2551487b675bc\nSuccessfully built gensim smart-open\nInstalling collected packages: boto, smart-open, gensim\nSuccessfully installed boto-2.49.0 gensim-3.8.2 smart-open-1.11.1\n" ], [ "# import glove embeddings into a word2vec format that is consumable by Gensim\nfrom gensim.scripts.glove2word2vec import glove2word2vec\nglove_input_file = 'glove.6B.100d.txt'\nword2vec_output_file = 'glove.6B.100d.txt.word2vec'\nglove2word2vec(glove_input_file, word2vec_output_file)", "_____no_output_____" ], [ "from gensim.models import KeyedVectors\n# load the Stanford GloVe model\nfilename = 'glove.6B.100d.txt.word2vec'\nmodel = KeyedVectors.load_word2vec_format(filename, binary=False)\n# calculate: (king - man) + woman = ?\nresult = model.most_similar(positive=['woman', 'king'], negative=['man'], topn=1)\nprint(result)", "[('queen', 0.7698541283607483)]\n" ], [ "words = [\"woman\", \"king\", \"man\", \"queen\", \"puppy\", \"kitten\", \"cat\", \n \"quarterback\", \"football\", \"stadium\", \"touchdown\",\n \"dog\", \"government\", \"tax\", \"federal\", \"judicial\", \"elections\",\n \"avocado\", \"tomato\", \"pear\", \"championship\", \"playoffs\"]\n\nvectors = [model.wv[word] for word in words]\n\nimport pandas as pd\nvector_df = pd.DataFrame(vectors)\nvector_df[\"word\"] = words\nvector_df.head()", "_____no_output_____" ] ], [ [ "## Using Spacy word2vec embeddings", "_____no_output_____" ] ], [ [ "import en_core_web_md\nimport spacy\nfrom scipy.spatial.distance import cosine\nnlp = en_core_web_md.load()", "_____no_output_____" ], [ "words = [\"woman\", \"king\", \"man\", \"queen\", \"puppy\", \"kitten\", \"cat\", \n \"quarterback\", \"football\", \"stadium\", \"touchdown\",\n \"dog\", \"government\", \"tax\", \"federal\", \"judicial\", \"elections\",\n \"avocado\", \"tomato\", \"pear\", \"championship\", \"playoffs\"]\n", "_____no_output_____" ], [ "tokens = nlp(\" \".join(words))\nword2vec_vectors = [token.vector for token in tokens]", "_____no_output_____" ], [ "np.array(word2vec_vectors).shape", "_____no_output_____" ], [ "%matplotlib inline\nfrom sklearn.manifold import TSNE\nfrom sklearn.decomposition import PCA\nfrom sklearn.decomposition import TruncatedSVD\nimport matplotlib.pyplot as plt\nimport matplotlib\ndimension_model = PCA(n_components=2)\nreduced_vectors = dimension_model.fit_transform(word2vec_vectors)", "_____no_output_____" ], [ "reduced_vectors.shape", "_____no_output_____" ], [ "matplotlib.rc('figure', figsize=(10, 10))\nfor i, vector in enumerate(reduced_vectors):\n x = vector[0]\n y = vector[1]\n plt.plot(x,y, 'bo')\n plt.text(x * (1 + 0.01), y * (1 + 0.01) , words[i], fontsize=12)", "_____no_output_____" ] ], [ [ "## Using Glove", "_____no_output_____" ] ], [ [ "%matplotlib inline\nfrom sklearn.manifold import TSNE\nfrom sklearn.decomposition import PCA\nfrom sklearn.decomposition import TruncatedSVD\nimport matplotlib.pyplot as plt\n\ndimension_model = PCA(n_components=2)\nreduced_vectors = dimension_model.fit_transform(vectors)\n\nfor i, vector in enumerate(reduced_vectors):\n x = vector[0]\n y = vector[1]\n plt.plot(x,y, 'bo')\n plt.text(x * (1 + 0.01), y * (1 + 0.01) , words[i], fontsize=12)\n\n", "_____no_output_____" ] ], [ [ "# Clustering Text", "_____no_output_____" ] ], [ [ "from sklearn.cluster import KMeans\n\nkmeans = KMeans(n_clusters=4)\ncluster_assignments = kmeans.fit_predict(reduced_vectors)\nfor cluster_assignment, word in zip(cluster_assignments, words):\n print(f\"{word} assigned to cluster {cluster_assignment}\")\n", "woman assigned to cluster 1\nking assigned to cluster 1\nman assigned to cluster 1\nqueen assigned to cluster 1\npuppy assigned to cluster 3\nkitten assigned to cluster 3\ncat assigned to cluster 3\nquarterback assigned to cluster 2\nfootball assigned to cluster 2\nstadium assigned to cluster 2\ntouchdown assigned to cluster 2\ndog assigned to cluster 3\ngovernment assigned to cluster 0\ntax assigned to cluster 0\nfederal assigned to cluster 0\njudicial assigned to cluster 0\nelections assigned to cluster 0\navocado assigned to cluster 3\ntomato assigned to cluster 3\npear assigned to cluster 3\nchampionship assigned to cluster 2\nplayoffs assigned to cluster 2\n" ], [ "color_map = {\n 0: \"r\",\n 1: \"b\",\n 2: \"g\",\n 3: \"y\"\n}\n\nplt.rcParams[\"figure.figsize\"] = (10,10)\n\nfor i, vector in enumerate(reduced_vectors):\n x = vector[0]\n y = vector[1]\n plt.plot(x,y, 'bo', c=color_map[cluster_assignments[i]])\n plt.text(x * (1 + 0.01), y * (1 + 0.01) , words[i], fontsize=12)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## 용어 정의", "_____no_output_____" ] ], [ [ "#가설설정\n# A hypothesis test is a statistical method that uses sample data to evaluate a hypothesis about a population.\n\n1. First, we state a hypothesis about a population. Usually the hypothesis concerns the value of a population parameter. \n2. Before we select a sample, we use the hypothesis to predict the characteristics that the sample should have. \n3. Next, we obtain a random sample from the population. \n4. Finally, we compare the obtained sample data with the prediction that was made from the hypothesis.\n\n## 가설설정 프로세스\n1. State the hypothesis. null hypothesis(H0) \n\n귀무가설 : 독립변수가 종속변수에 어떤 영향을 미치지 않는다는 것 => 레스토랑의 웨이터가 레드 셔츠 입는 것이 팁에 영향이 없다.\nThe null hypothesis (H0) states that in the general population \nthere is no change, no difference, or no relationship.\nIn the context of an experiment, \nH0 predicts that the independent variable (treatment) \nhas no effect on the dependent variable (scores) for the population.\n\nm = 15.8\n\n대안가설 : 어떤 변인이 종속 변수에 효과가 있다는 것 => 레스토랑의 웨이터가 레드 셔츠 입는 것 팁에 영향이 있다. \nThe alternative hypothesis (H1) states that there is a change, a difference, \nor a relationship for the general population. \nIn the context of an experiment,\nH1 predicts that the independent variable (treatment) does have an effect on the dependent variable.\n \nm != 15.8 이다.\n\n이 실험에서는 \nm > 15.8\n\ndirectional hypothisis test\n\n2. set the criteria for a decision\n\na. Sample means that are likely to be obtained if H0 is true; \nthat is, sample means that are close to the null hypothesis\n\nb. Sample means that are very unlikely to be obtained if H0 is true;\nthat is, sample means that are very different from the null hypothesis\n\n\nThe Alpha Level\n\nalpha levels are α = .05 (5%), α = .01 (1%), and α = .001 (0.1%).\n\nThe alpha level, or the level of significance, \nis a probability value that is used to define the concept of \n“very unlikely” in a hypothesis test.\n\nThe critical region is composed of the extreme sample values that are very unlikely (as defined by the alpha level) to be obtained if the null hypothesis is true. The boundaries for the critical region are determined by the alpha level. \nIf sample data fall in the critical region, the null hypothesis is rejected.\n\n\n3. Collect data and compute sample statistics.\n\nz = sample mean - hypothesized population mean / standard error between M and m\n\n4. Make a decision\n\n1. Thesampledataarelocatedinthecriticalregion.\nBydefinition,asamplevaluein the critical region is very unlikely to occur if the null hypothesis is true. \n\n2. The sample data are not in the critical region. \nIn this case, the sample mean is reasonably close to the population mean specified in the null hypothesis (in the center of the distribution).", "_____no_output_____" ] ], [ [ "# Problems", "_____no_output_____" ] ], [ [ "1. Identify the four steps of a hypothesis test as presented in this chapter.\n\n1)State the hypothesis.\n귀무가설과 대안가설 언급\n\n2)alpha level 설정, 신뢰 구간 설정\n\n3) Collect data and compute sample statistics. \n데이터 수집과 샘플 통계적 계산\n\n4)make decision\n결론 결정\n", "_____no_output_____" ], [ "2. Define the alpha level and the critical region for a hypothesis test.\n독립변수와 종속변수에 대한 귀무가설을 reject하기 위해 그 통계치를 통상적인 수치를 벗어나 의미있는 수가 나온 것을 설정해준다. ", "_____no_output_____" ], [ "3. Define a Type I error and a Type II error and explain the consequences of each.\n가설검증에서 실제효과가 없는데 효과가 있는 것으로 나온것, 실제 효과가 있는데, 없는 것으로 나온것. 가설 설정에 문제", "_____no_output_____" ], [ "4. If the alpha level is changed from α = .05 to α = .01,\na. What happens to the boundaries for the critical\nregion?\n신뢰구간이 줄어든다.\nb. What happens to the probability of a Type I error?\n에러 확률은 낮아진다. ", "_____no_output_____" ], [ "6. Although there is a popular belief that herbal remedies such as Ginkgo biloba and Ginseng may improve learning and memory in healthy adults, these effects are usually not supported by well- controlled research (Persson, Bringlov, Nilsson, and Nyberg, 2004). In a typical study, a researcher\nobtains a sample of n = 16 participants and has each person take the herbal supplements every day for\n90 days. At the end of the 90 days, each person takes a standardized memory test. For the general popula- tion, scores from the test form a normal distribution with a mean of μ = 50 and a standard deviation of\nσ = 12. The sample of research participants had an average of M = 54.\na. Assuming a two-tailed test, state the null hypoth-\nesis in a sentence that includes the two variables\nbeing examined.\nb. Using the standard 4-step procedure, conduct a\ntwo-tailed hypothesis test with α = .05 to evaluate the effect of the supplements.\n\n\n", "_____no_output_____" ], [ "from scipy import stats\n\nsample_number = 16 # 샘플수 \npopulation_mean = 50 # 모집단의 평균\nstandard_deviation = 12 # 표준편차\nsample_mean = 54 # 샘플의 평균\n\n\nresult = stats.ttest_1samp(sample_mean, 50) # 비교집단, 관측치\n\nresult", "_____no_output_____" ], [ "sample_mean - population_mean", "_____no_output_____" ], [ "## Import \nimport numpy as np\nfrom scipy import stats\n\n\nsample_number = 16 # 샘플수 \npopulation_mean = 50 # 모집단의 평균\nstandard_deviation = 12 # 표준편차\nsample_mean = 54 # 샘플의 평균\n\n\n## 신뢰구간을 벗어나는지 아닌지 확인 함수 \n\nalpha_level05 = 1.96\nalpha_level01 = 2.58\n\ndef h_test(sample_mean, population_mean, standard_deviation, sample_number, alpha_level):\n \n \n result = (sample_mean - population_mean)/ (standard_deviation/np.sqrt(sample_number))\n \n \n if result> alpha_level or result< - alpha_level:\n print(\"a = .05 신뢰구간에서 귀무가설 reject되고, 가설이 ok\")\n\n else:\n print(\"귀무가설이 reject 되지 않아 가설이 기각됩니다.\")\n\n return result\n\n\n##Compute Cohen’s d\ndef Cohen(sample_mean, population_mean, standard_deviation):\n result = (sample_mean - population_mean) / (standard_deviation)\n \n \n if result<=0.2:\n print(\"small effect\")\n\n elif result<= 0.5:\n print(\"medium effect\")\n\n elif result<= 0.8:\n print(\"Large effect\")\n \n return result\n\n ", "_____no_output_____" ], [ "## 신뢰구간을 벗어나는지 아닌지 확인 함수 \nh_test(sample_mean, population_mean, standard_deviation, sample_number, alpha_level05)\n\n \n", "귀무가설이 reject 되지 않아 가설이 기각됩니다.\n" ], [ "Cohen(sample_mean, population_mean, standard_deviation)", "medium effect\n" ], [ "함수를 활용해서, 신뢰구간과 cohen's d를 구할 수 있다. \n", "_____no_output_____" ], [ "# ## Import the packages\n# import numpy as np\n# from scipy import stats\n\n\n# ## 함수로 만들기\n\n# #Sample Size\n# sample_number = 16 \n# population_mean = 50 # 모집단의 평균\n# standard_deviation = 12 # 표준편차\n# sample_mean = [54,54,58,53,52] # 샘플의 평균\n\n# def h_test(sample_mean, population_mean, standard_deviation, sample_number):\n \n# #For unbiased max likelihood estimate we have to divide the var by N-1, and therefore the parameter ddof = 1\n# var_sample_mean = sample_mean.var(ddof=1)\n# var_population_mean = population_mean.var(ddof=1)\n \n# #std deviation\n# std_deviation = np.sqrt((var_sample_mean + var_population_mean)/2)\n \n# ## Calculate the t-statistics\n# t = (a.mean() - b.mean())/(s*np.sqrt(2/N))\n \n# ## Define 2 random distributions\n\n# N = 10\n# #Gaussian distributed data with mean = 2 and var = 1\n# a = np.random.randn(N) + 2\n# #Gaussian distributed data with with mean = 0 and var = 1\n# b = np.random.randn(N)\n\n\n# ## Calculate the Standard Deviation\n# #Calculate the variance to get the standard deviation\n\n# #For unbiased max likelihood estimate we have to divide the var by N-1, and therefore the parameter ddof = 1\n# var_a = a.var(ddof=1)\n# var_b = b.var(ddof=1)\n\n# #std deviation\n# s = np.sqrt((var_a + var_b)/2)\n# s\n\n\n\n# ## Calculate the t-statistics\n# t = (a.mean() - b.mean())/(s*np.sqrt(2/N))\n\n\n\n# ## Compare with the critical t-value\n# #Degrees of freedom\n# df = 2*N - 2\n\n# #p-value after comparison with the t \n# p = 1 - stats.t.cdf(t,df=df)\n\n\n# print(\"t = \" + str(t))\n# print(\"p = \" + str(2*p))\n# ### You can see that after comparing the t statistic with the critical t value (computed internally) we get a good p value of 0.0005 and thus we reject the null hypothesis and thus it proves that the mean of the two distributions are different and statistically significant.\n\n\n# ## Cross Checking with the internal scipy function\n# t2, p2 = stats.ttest_ind(a,b)\n# print(\"t = \" + str(t2))\n# print(\"p = \" + str(p2))", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "Source: https://qiskit.org/documentation/tutorials/circuits/01_circuit_basics.html", "_____no_output_____" ], [ "## Circuit Basics", "_____no_output_____" ] ], [ [ "import numpy as np\nfrom qiskit import QuantumCircuit\n\n%matplotlib inline", "_____no_output_____" ] ], [ [ "Create a Quantum Circuit acting on a quantum register of three qubits", "_____no_output_____" ] ], [ [ "circ = QuantumCircuit(3)", "_____no_output_____" ] ], [ [ "After you create the circuit with its registers, you can add gates (“operations”) to manipulate the registers. As you proceed through the tutorials you will find more gates and circuits; below is an example of a quantum circuit that makes a three-qubit GHZ state\n\n|𝜓⟩=(|000⟩+|111⟩)/2‾√\n\nTo create such a state, we start with a three-qubit quantum register. By default, each qubit in the register is initialized to |0⟩. To make the GHZ state, we apply the following gates: - A Hadamard gate 𝐻 on qubit 0, which puts it into the superposition state (|0⟩+|1⟩)/2‾√. - A controlled-Not operation (𝐶𝑋) between qubit 0 and qubit 1. - A controlled-Not operation between qubit 0 and qubit 2.\n\nOn an ideal quantum computer, the state produced by running this circuit would be the GHZ state above.", "_____no_output_____" ] ], [ [ "# Add a H gate on qubit 0, putting this qubit in superposition.\ncirc.h(0)\n# Add a CX (CNOT) gate on control qubit 0 and target qubit 1, putting\n# the qubits in a Bell state.\ncirc.cx(0, 1)\n# Add a CX (CNOT) gate on control qubit 0 and target qubit 2, putting\n# the qubits in a GHZ state.\ncirc.cx(0, 2)", "_____no_output_____" ] ], [ [ "Visualize circuit", "_____no_output_____" ] ], [ [ "circ.draw('mpl')", "_____no_output_____" ] ], [ [ "Simulating Circuits", "_____no_output_____" ], [ "To simulate a circuit we use the quant-info module in Qiskit. This simulator returns the quantum state, which is a complex vector of dimensions 2𝑛, where 𝑛 is the number of qubits (so be careful using this as it will quickly get too large to run on your machine).\n\nThere are two stages to the simulator. The fist is to set the input state and the second to evolve the state by the quantum circuit.", "_____no_output_____" ] ], [ [ "from qiskit.quantum_info import Statevector\n\n# Set the intial state of the simulator to the ground state using from_int\nstate = Statevector.from_int(0, 2**3)\n\n# Evolve the state by the quantum circuit\nstate = state.evolve(circ)\n\n#draw using latex\nstate.draw('latex')", "_____no_output_____" ] ], [ [ "Visualization\n\nBelow, we use the visualization function to plot the qsphere and a hinton representing the real and imaginary components of the state density matrix 𝜌.", "_____no_output_____" ] ], [ [ "state.draw('qsphere')", "/usr/local/lib/python3.9/site-packages/qiskit/visualization/state_visualization.py:705: MatplotlibDeprecationWarning: \nThe M attribute was deprecated in Matplotlib 3.4 and will be removed two minor releases later. Use self.axes.M instead.\n xs, ys, _ = proj3d.proj_transform(xs3d, ys3d, zs3d, renderer.M)\n" ], [ "state.draw('hinton')", "_____no_output_____" ] ], [ [ "Unitary representation of a circuit\n\nQiskit’s quant_info module also has an operator method which can be used to make a unitary operator for the circuit. This calculates the 2𝑛×2𝑛 matrix representing the quantum circuit.", "_____no_output_____" ] ], [ [ "from qiskit.quantum_info import Operator\n\nU = Operator(circ)\n\n# Show the results\nU.data", "_____no_output_____" ] ], [ [ "OpenQASM backend\n\nThe simulators above are useful because they provide information about the state output by the ideal circuit and the matrix representation of the circuit. However, a real experiment terminates by measuring each qubit (usually in the computational |0⟩,|1⟩ basis). Without measurement, we cannot gain information about the state. Measurements cause the quantum system to collapse into classical bits.\n\nFor example, suppose we make independent measurements on each qubit of the three-qubit GHZ state\n\n|𝜓⟩=(|000⟩+|111⟩)/2‾√,\n\nand let 𝑥𝑦𝑧 denote the bitstring that results. Recall that, under the qubit labeling used by Qiskit, 𝑥 would correspond to the outcome on qubit 2, 𝑦 to the outcome on qubit 1, and 𝑧 to the outcome on qubit 0.\n\nNote: This representation of the bitstring puts the most significant bit (MSB) on the left, and the least significant bit (LSB) on the right. This is the standard ordering of binary bitstrings. We order the qubits in the same way (qubit representing the MSB has index 0), which is why Qiskit uses a non-standard tensor product order.\n\nRecall the probability of obtaining outcome 𝑥𝑦𝑧 is given by\n\nPr(𝑥𝑦𝑧)=|⟨𝑥𝑦𝑧|𝜓⟩|2\n\nand as such for the GHZ state probability of obtaining 000 or 111 are both 1/2.\n\nTo simulate a circuit that includes measurement, we need to add measurements to the original circuit above, and use a different Aer backend.", "_____no_output_____" ] ], [ [ "# Create a Quantum Circuit\nmeas = QuantumCircuit(3, 3)\nmeas.barrier(range(3))\n# map the quantum measurement to the classical bits\nmeas.measure(range(3), range(3))\n\n# The Qiskit circuit object supports composition.\n# Here the meas has to be first and front=True (putting it before)\n# as compose must put a smaller circuit into a larger one.\nqc = meas.compose(circ, range(3), front=True)\n\n#drawing the circuit\nqc.draw('mpl')", "_____no_output_____" ] ], [ [ "This circuit adds a classical register, and three measurements that are used to map the outcome of qubits to the classical bits.\n\nTo simulate this circuit, we use the qasm_simulator in Qiskit Aer. Each run of this circuit will yield either the bitstring 000 or 111. To build up statistics about the distribution of the bitstrings (to, e.g., estimate Pr(000)), we need to repeat the circuit many times. The number of times the circuit is repeated can be specified in the execute function, via the shots keyword.", "_____no_output_____" ] ], [ [ "# Adding the transpiler to reduce the circuit to QASM instructions\n# supported by the backend\nfrom qiskit import transpile\n\n# Use Aer's qasm_simulator\nfrom qiskit.providers.aer import QasmSimulator\n\nbackend = QasmSimulator()\n\n# First we have to transpile the quantum circuit\n# to the low-level QASM instructions used by the\n# backend\nqc_compiled = transpile(qc, backend)\n\n# Execute the circuit on the qasm simulator.\n# We've set the number of repeats of the circuit\n# to be 1024, which is the default.\njob_sim = backend.run(qc_compiled, shots=1024)\n\n# Grab the results from the job.\nresult_sim = job_sim.result()", "_____no_output_____" ] ], [ [ "\nOnce you have a result object, you can access the counts via the function get_counts(circuit). This gives you the aggregated binary outcomes of the circuit you submitted.", "_____no_output_____" ] ], [ [ "counts = result_sim.get_counts(qc_compiled)\nprint(counts)", "{'000': 503, '111': 521}\n" ] ], [ [ "Approximately 50 percent of the time, the output bitstring is 000. Qiskit also provides a function plot_histogram, which allows you to view the outcomes.", "_____no_output_____" ] ], [ [ "from qiskit.visualization import plot_histogram\nplot_histogram(counts)", "_____no_output_____" ] ], [ [ "The estimated outcome probabilities Pr(000) and Pr(111) are computed by taking the aggregate counts and dividing by the number of shots (times the circuit was repeated). Try changing the shots keyword in the execute function and see how the estimated probabilities change.", "_____no_output_____" ] ], [ [ "import qiskit.tools.jupyter\n%qiskit_version_table\n%qiskit_copyright", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Linear algebra", "_____no_output_____" ] ], [ [ "import numpy as np", "_____no_output_____" ], [ "np.__version__", "_____no_output_____" ] ], [ [ "## Matrix and vector products", "_____no_output_____" ], [ "Q1. Predict the results of the following code.", "_____no_output_____" ] ], [ [ "import numpy as np\nx = [1,2]\ny = [[4, 1], [2, 2]]\nprint np.dot(x, y)\nprint np.dot(y, x)\nprint np.matmul(x, y)\nprint np.inner(x, y)\nprint np.inner(y, x)", "[8 5]\n[6 6]\n[8 5]\n[6 6]\n[6 6]\n" ] ], [ [ "Q2. Predict the results of the following code.", "_____no_output_____" ] ], [ [ "x = [[1, 0], [0, 1]]\ny = [[4, 1], [2, 2], [1, 1]]\nprint np.dot(y, x)\nprint np.matmul(y, x)\n", "[[4 1]\n [2 2]\n [1 1]]\n[[4 1]\n [2 2]\n [1 1]]\n" ] ], [ [ "Q3. Predict the results of the following code.", "_____no_output_____" ] ], [ [ "x = np.array([[1, 4], [5, 6]])\ny = np.array([[4, 1], [2, 2]])\nprint np.vdot(x, y)\nprint np.vdot(y, x)\nprint np.dot(x.flatten(), y.flatten())\nprint np.inner(x.flatten(), y.flatten())\nprint (x*y).sum()", "30\n30\n30\n30\n30\n" ] ], [ [ "Q4. Predict the results of the following code.", "_____no_output_____" ] ], [ [ "x = np.array(['a', 'b'], dtype=object)\ny = np.array([1, 2])\nprint np.inner(x, y)\nprint np.inner(y, x)\nprint np.outer(x, y)\nprint np.outer(y, x)", "abb\nabb\n[['a' 'aa']\n ['b' 'bb']]\n[['a' 'b']\n ['aa' 'bb']]\n" ] ], [ [ "## Decompositions", "_____no_output_____" ], [ "Q5. Get the lower-trianglular `L` in the Cholesky decomposition of x and verify it.", "_____no_output_____" ] ], [ [ "x = np.array([[4, 12, -16], [12, 37, -43], [-16, -43, 98]], dtype=np.int32)\nL = np.linalg.cholesky(x)\nprint L\nassert np.array_equal(np.dot(L, L.T.conjugate()), x)", "[[ 2. 0. 0.]\n [ 6. 1. 0.]\n [-8. 5. 3.]]\n" ] ], [ [ "Q6. Compute the qr factorization of x and verify it.", "_____no_output_____" ] ], [ [ "x = np.array([[12, -51, 4], [6, 167, -68], [-4, 24, -41]], dtype=np.float32)\nq, r = np.linalg.qr(x)\nprint \"q=\\n\", q, \"\\nr=\\n\", r\nassert np.allclose(np.dot(q, r), x)", "q=\n[[-0.85714287 0.39428571 0.33142856]\n [-0.42857143 -0.90285712 -0.03428571]\n [ 0.2857143 -0.17142858 0.94285715]] \nr=\n[[ -14. -21. 14.]\n [ 0. -175. 70.]\n [ 0. 0. -35.]]\n" ] ], [ [ "Q7. Factor x by Singular Value Decomposition and verify it.", "_____no_output_____" ] ], [ [ "x = np.array([[1, 0, 0, 0, 2], [0, 0, 3, 0, 0], [0, 0, 0, 0, 0], [0, 2, 0, 0, 0]], dtype=np.float32)\nU, s, V = np.linalg.svd(x, full_matrices=False)\nprint \"U=\\n\", U, \"\\ns=\\n\", s, \"\\nV=\\n\", v\nassert np.allclose(np.dot(U, np.dot(np.diag(s), V)), x)\n", "U=\n[[ 0. 1. 0. 0.]\n [ 1. 0. 0. 0.]\n [ 0. 0. 0. -1.]\n [ 0. 0. 1. 0.]] \ns=\n[ 3. 2.23606801 2. 0. ] \nV=\n[[ 1. 0. 0.]\n [ 0. 1. 0.]\n [ 0. 0. 1.]]\n" ] ], [ [ "## Matrix eigenvalues", "_____no_output_____" ], [ "Q8. Compute the eigenvalues and right eigenvectors of x. (Name them eigenvals and eigenvecs, respectively)", "_____no_output_____" ] ], [ [ "x = np.diag((1, 2, 3))\neigenvals = np.linalg.eig(x)[0]\neigenvals_ = np.linalg.eigvals(x)\nassert np.array_equal(eigenvals, eigenvals_)\nprint \"eigenvalues are\\n\", eigenvals\neigenvecs = np.linalg.eig(x)[1]\nprint \"eigenvectors are\\n\", eigenvecs", "eigenvalues are\n[ 1. 2. 3.]\neigenvectors are\n[[ 1. 0. 0.]\n [ 0. 1. 0.]\n [ 0. 0. 1.]]\n" ] ], [ [ "Q9. Predict the results of the following code.", "_____no_output_____" ] ], [ [ "print np.array_equal(np.dot(x, eigenvecs), eigenvals * eigenvecs)", "True\n" ] ], [ [ "## Norms and other numbers", "_____no_output_____" ], [ "Q10. Calculate the Frobenius norm and the condition number of x.", "_____no_output_____" ] ], [ [ "x = np.arange(1, 10).reshape((3, 3))\nprint np.linalg.norm(x, 'fro')\nprint np.linalg.cond(x, 'fro')", "16.8819430161\n4.56177073661e+17\n" ] ], [ [ "Q11. Calculate the determinant of x.", "_____no_output_____" ] ], [ [ "x = np.arange(1, 5).reshape((2, 2))\nout1 = np.linalg.det(x)\nout2 = x[0, 0] * x[1, 1] - x[0, 1] * x[1, 0]\nassert np.allclose(out1, out2)\nprint out1", "-2.0\n" ] ], [ [ "Q12. Calculate the rank of x.", "_____no_output_____" ] ], [ [ "x = np.eye(4)\nout1 = np.linalg.matrix_rank(x)\nout2 = np.linalg.svd(x)[1].size\nassert out1 == out2\nprint out1\n", "4\n" ] ], [ [ "Q13. Compute the sign and natural logarithm of the determinant of x.", "_____no_output_____" ] ], [ [ "x = np.arange(1, 5).reshape((2, 2))\nsign, logdet = np.linalg.slogdet(x)\ndet = np.linalg.det(x)\nassert sign == np.sign(det)\nassert logdet == np.log(np.abs(det))\nprint sign, logdet\n", "-1.0 0.69314718056\n" ] ], [ [ "Q14. Return the sum along the diagonal of x.", "_____no_output_____" ] ], [ [ "x = np.eye(4)\nout1 = np.trace(x)\nout2 = x.diagonal().sum()\nassert out1 == out2\nprint out1", "4.0\n" ] ], [ [ "## Solving equations and inverting matrices", "_____no_output_____" ], [ "Q15. Compute the inverse of x.", "_____no_output_____" ] ], [ [ "x = np.array([[1., 2.], [3., 4.]])\nout1 = np.linalg.inv(x)\nassert np.allclose(np.dot(x, out1), np.eye(2))\nprint out1\n", "[[-2. 1. ]\n [ 1.5 -0.5]]\n" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# 自然语言处理实战——命名实体识别\n\n### 进入ModelArts\n\n点击如下链接：https://www.huaweicloud.com/product/modelarts.html ， 进入ModelArts主页。点击“立即使用”按钮，输入用户名和密码登录，进入ModelArts使用页面。\n\n### 创建ModelArts notebook\n\n下面，我们在ModelArts中创建一个notebook开发环境，ModelArts notebook提供网页版的Python开发环境，可以方便的编写、运行代码，并查看运行结果。\n\n第一步：在ModelArts服务主界面依次点击“开发环境”、“创建”\n\n\n\n第二步：填写notebook所需的参数：\n\n| 参数 | 说明 |\n| - - - - - | - - - - - |\n| 计费方式 | 按需计费 |\n| 名称 | Notebook实例名称 |\n| 工作环境 | Python3 |\n| 资源池 | 选择\"公共资源池\"即可 |\n| 类型 | 选择\"GPU\" |\n| 规格 | 选择\"[限时免费]体验规格GPU版\"|\n| 存储配置 | 选择EVS，磁盘规格5GB |\n\n第三步：配置好notebook参数后，点击下一步，进入notebook信息预览。确认无误后，点击“立即创建”\n\n第四步：创建完成后，返回开发环境主界面，等待Notebook创建完毕后，打开Notebook，进行下一步操作。\n\n\n### 在ModelArts中创建开发环境\n\n接下来，我们创建一个实际的开发环境，用于后续的实验步骤。\n\n第一步：点击下图所示的“打开”按钮，进入刚刚创建的Notebook\n\n\n第二步：创建一个Python3环境的的Notebook。点击右上角的\"New\"，然后创建TensorFlow 1.13.1开发环境。\n\n第三步：点击左上方的文件名\"Untitled\"，并输入一个与本实验相关的名称\n\n\n\n\n\n### 在Notebook中编写并执行代码\n\n在Notebook中，我们输入一个简单的打印语句，然后点击上方的运行按钮，可以查看语句执行的结果：\n\n\n\n开发环境准备好啦，接下来可以愉快地写代码啦！\n\n\n### 准备源代码和数据\n\n准备案例所需的源代码和数据，相关资源已经保存在 OBS 中，我们通过 ModelArts SDK 将资源下载到本地。", "_____no_output_____" ] ], [ [ "from modelarts.session import Session\nsession = Session()\n\nif session.region_name == 'cn-north-1':\n bucket_path = 'modelarts-labs/notebook/DL_nlp_ner/ner.tar.gz'\n \nelif session.region_name == 'cn-north-4':\n bucket_path = 'modelarts-labs-bj4/notebook/DL_nlp_ner/ner.tar.gz'\nelse:\n print(\"请更换地区到北京一或北京四\")\n \nsession.download_data(bucket_path=bucket_path, path='./ner.tar.gz')\n\n!ls -la ", "Successfully download file modelarts-labs/notebook/DL_nlp_ner/ner.tar.gz from OBS to local ./ner.tar.gz\ntotal 375220\r\ndrwxrwsrwx 4 ma-user ma-group 4096 Sep 6 13:34 .\r\ndrwsrwsr-x 22 ma-user ma-group 4096 Sep 6 13:03 ..\r\ndrwxr-s--- 2 ma-user ma-group 4096 Sep 6 13:33 .ipynb_checkpoints\r\n-rw-r----- 1 ma-user ma-group 45114 Sep 6 13:33 ner.ipynb\r\n-rw-r----- 1 ma-user ma-group 384157325 Sep 6 13:35 ner.tar.gz\r\ndrwx--S--- 2 ma-user ma-group 4096 Sep 6 13:03 .Trash-1000\r\n" ] ], [ [ "解压从obs下载的压缩包，解压后删除压缩包。", "_____no_output_____" ] ], [ [ "# 解压\n!tar xf ./ner.tar.gz\n\n# 删除\n!rm ./ner.tar.gz\n\n!ls -la ", "total 68\r\ndrwxrwsrwx 5 ma-user ma-group 4096 Sep 6 13:35 .\r\ndrwsrwsr-x 22 ma-user ma-group 4096 Sep 6 13:03 ..\r\ndrwxr-s--- 2 ma-user ma-group 4096 Sep 6 13:33 .ipynb_checkpoints\r\ndrwxr-s--- 8 ma-user ma-group 4096 Sep 6 00:24 ner\r\n-rw-r----- 1 ma-user ma-group 45114 Sep 6 13:33 ner.ipynb\r\ndrwx--S--- 2 ma-user ma-group 4096 Sep 6 13:03 .Trash-1000\r\n" ] ], [ [ "#### 导入Python库", "_____no_output_____" ] ], [ [ "import os\nimport json\nimport numpy as np\nimport tensorflow as tf\nimport codecs\nimport pickle\nimport collections\nfrom ner.bert import modeling, optimization, tokenization", "_____no_output_____" ] ], [ [ "#### 定义路径及参数", "_____no_output_____" ] ], [ [ "data_dir = \"./ner/data\" \noutput_dir = \"./ner/output\" \nvocab_file = \"./ner/chinese_L-12_H-768_A-12/vocab.txt\" \ndata_config_path = \"./ner/chinese_L-12_H-768_A-12/bert_config.json\" \ninit_checkpoint = \"./ner/chinese_L-12_H-768_A-12/bert_model.ckpt\" \nmax_seq_length = 128 \nbatch_size = 64 \nnum_train_epochs = 5.0 ", "_____no_output_____" ] ], [ [ "#### 定义processor类获取数据，打印标签", "_____no_output_____" ] ], [ [ "tf.logging.set_verbosity(tf.logging.INFO)\nfrom ner.src.models import InputFeatures, InputExample, DataProcessor, NerProcessor\n\nprocessors = {\"ner\": NerProcessor }\nprocessor = processors[\"ner\"](output_dir)\n\nlabel_list = processor.get_labels()\nprint(\"labels:\", label_list)", "labels: ['O', 'B-PER', 'I-PER', 'B-ORG', 'I-ORG', 'B-LOC', 'I-LOC', 'X', '[CLS]', '[SEP]']\n" ] ], [ [ "以上 labels 分别表示：\n\n- O：非标注实体\n- B-PER：人名首字\n- I-PER：人名非首字\n- B-ORG：组织首字\n- I-ORG：组织名非首字\n- B-LOC：地名首字\n- I-LOC：地名非首字\n- X：未知\n- [CLS]：句首\n- [SEP]：句尾\n\n#### 加载预训练参数", "_____no_output_____" ] ], [ [ "data_config = json.load(codecs.open(data_config_path))\ntrain_examples = processor.get_train_examples(data_dir) \nnum_train_steps = int(len(train_examples) / batch_size * num_train_epochs) \nnum_warmup_steps = int(num_train_steps * 0.1) \ndata_config['num_train_steps'] = num_train_steps\ndata_config['num_warmup_steps'] = num_warmup_steps\ndata_config['num_train_size'] = len(train_examples)\n\nprint(\"显示配置信息:\")\nfor key,value in data_config.items():\n print('{key}:{value}'.format(key = key, value = value))\n\nbert_config = modeling.BertConfig.from_json_file(data_config_path)\ntokenizer = tokenization.FullTokenizer(vocab_file=vocab_file, do_lower_case=True)\n\n#tf.estimator运行参数\nrun_config = tf.estimator.RunConfig(\n model_dir=output_dir,\n save_summary_steps=1000,\n save_checkpoints_steps=1000,\n session_config=tf.ConfigProto(\n log_device_placement=False,\n inter_op_parallelism_threads=0,\n intra_op_parallelism_threads=0,\n allow_soft_placement=True\n )\n)", "显示配置信息:\nattention_probs_dropout_prob:0.1\ndirectionality:bidi\nhidden_act:gelu\nhidden_dropout_prob:0.1\nhidden_size:768\ninitializer_range:0.02\nintermediate_size:3072\nmax_position_embeddings:512\nnum_attention_heads:12\nnum_hidden_layers:12\npooler_fc_size:768\npooler_num_attention_heads:12\npooler_num_fc_layers:3\npooler_size_per_head:128\npooler_type:first_token_transform\ntype_vocab_size:2\nvocab_size:21128\nnum_train_steps:1630\nnum_warmup_steps:163\nnum_train_size:20864\n" ] ], [ [ "#### 读取数据，获取句向量", "_____no_output_____" ] ], [ [ "def convert_single_example(ex_index, example, label_list, max_seq_length, \n tokenizer, output_dir, mode):\n label_map = {}\n for (i, label) in enumerate(label_list, 1):\n label_map[label] = i\n if not os.path.exists(os.path.join(output_dir, 'label2id.pkl')):\n with codecs.open(os.path.join(output_dir, 'label2id.pkl'), 'wb') as w:\n pickle.dump(label_map, w)\n\n textlist = example.text.split(' ')\n labellist = example.label.split(' ')\n tokens = []\n labels = []\n for i, word in enumerate(textlist):\n token = tokenizer.tokenize(word)\n tokens.extend(token)\n label_1 = labellist[i]\n for m in range(len(token)):\n if m == 0:\n labels.append(label_1)\n else: \n labels.append(\"X\")\n if len(tokens) >= max_seq_length - 1:\n tokens = tokens[0:(max_seq_length - 2)]\n labels = labels[0:(max_seq_length - 2)]\n ntokens = []\n segment_ids = []\n label_ids = []\n ntokens.append(\"[CLS]\") # 句子开始设置 [CLS] 标志\n segment_ids.append(0)\n label_ids.append(label_map[\"[CLS]\"]) \n for i, token in enumerate(tokens):\n ntokens.append(token)\n segment_ids.append(0)\n label_ids.append(label_map[labels[i]])\n ntokens.append(\"[SEP]\") # 句尾添加 [SEP] 标志\n segment_ids.append(0)\n label_ids.append(label_map[\"[SEP]\"])\n input_ids = tokenizer.convert_tokens_to_ids(ntokens) \n input_mask = [1] * len(input_ids)\n\n while len(input_ids) < max_seq_length:\n input_ids.append(0)\n input_mask.append(0)\n segment_ids.append(0)\n label_ids.append(0)\n ntokens.append(\"**NULL**\")\n\n assert len(input_ids) == max_seq_length\n assert len(input_mask) == max_seq_length\n assert len(segment_ids) == max_seq_length\n assert len(label_ids) == max_seq_length\n\n feature = InputFeatures(\n input_ids=input_ids,\n input_mask=input_mask,\n segment_ids=segment_ids,\n label_ids=label_ids,\n )\n \n return feature\n\ndef filed_based_convert_examples_to_features(\n examples, label_list, max_seq_length, tokenizer, output_file, mode=None):\n writer = tf.python_io.TFRecordWriter(output_file)\n for (ex_index, example) in enumerate(examples):\n if ex_index % 5000 == 0:\n tf.logging.info(\"Writing example %d of %d\" % (ex_index, len(examples)))\n feature = convert_single_example(ex_index, example, label_list, max_seq_length, tokenizer, output_dir, mode)\n\n def create_int_feature(values):\n f = tf.train.Feature(int64_list=tf.train.Int64List(value=list(values)))\n return f\n\n features = collections.OrderedDict()\n features[\"input_ids\"] = create_int_feature(feature.input_ids)\n features[\"input_mask\"] = create_int_feature(feature.input_mask)\n features[\"segment_ids\"] = create_int_feature(feature.segment_ids)\n features[\"label_ids\"] = create_int_feature(feature.label_ids)\n tf_example = tf.train.Example(features=tf.train.Features(feature=features))\n writer.write(tf_example.SerializeToString())\n\ntrain_file = os.path.join(output_dir, \"train.tf_record\")\n\n#将训练集中字符转化为features作为训练的输入\nfiled_based_convert_examples_to_features(\n train_examples, label_list, max_seq_length, tokenizer, output_file=train_file)", "INFO:tensorflow:Writing example 0 of 20864\nINFO:tensorflow:Writing example 5000 of 20864\nINFO:tensorflow:Writing example 10000 of 20864\nINFO:tensorflow:Writing example 15000 of 20864\nINFO:tensorflow:Writing example 20000 of 20864\n" ] ], [ [ "#### 引入 BiLSTM+CRF 层，作为下游模型", "_____no_output_____" ] ], [ [ "learning_rate = 5e-5 \ndropout_rate = 1.0 \nlstm_size=1 \ncell='lstm'\nnum_layers=1\n\nfrom ner.src.models import BLSTM_CRF\nfrom tensorflow.contrib.layers.python.layers import initializers\n\ndef create_model(bert_config, is_training, input_ids, input_mask,\n segment_ids, labels, num_labels, use_one_hot_embeddings,\n dropout_rate=dropout_rate, lstm_size=1, cell='lstm', num_layers=1):\n model = modeling.BertModel(\n config=bert_config,\n is_training=is_training,\n input_ids=input_ids,\n input_mask=input_mask,\n token_type_ids=segment_ids,\n use_one_hot_embeddings=use_one_hot_embeddings\n )\n embedding = model.get_sequence_output()\n max_seq_length = embedding.shape[1].value\n used = tf.sign(tf.abs(input_ids))\n lengths = tf.reduce_sum(used, reduction_indices=1) \n blstm_crf = BLSTM_CRF(embedded_chars=embedding, hidden_unit=1, cell_type='lstm', num_layers=1,\n dropout_rate=dropout_rate, initializers=initializers, num_labels=num_labels,\n seq_length=max_seq_length, labels=labels, lengths=lengths, is_training=is_training)\n rst = blstm_crf.add_blstm_crf_layer(crf_only=True)\n return rst\n\ndef model_fn_builder(bert_config, num_labels, init_checkpoint, learning_rate,\n num_train_steps, num_warmup_steps,use_one_hot_embeddings=False):\n #构建模型\n def model_fn(features, labels, mode, params):\n tf.logging.info(\"*** Features ***\")\n for name in sorted(features.keys()):\n tf.logging.info(\" name = %s, shape = %s\" % (name, features[name].shape))\n input_ids = features[\"input_ids\"]\n input_mask = features[\"input_mask\"]\n segment_ids = features[\"segment_ids\"]\n label_ids = features[\"label_ids\"]\n\n print('shape of input_ids', input_ids.shape)\n is_training = (mode == tf.estimator.ModeKeys.TRAIN)\n\n total_loss, logits, trans, pred_ids = create_model(\n bert_config, is_training, input_ids, input_mask, segment_ids, label_ids,\n num_labels, False, dropout_rate, lstm_size, cell, num_layers)\n\n tvars = tf.trainable_variables()\n\n if init_checkpoint:\n (assignment_map, initialized_variable_names) = \\\n modeling.get_assignment_map_from_checkpoint(tvars, init_checkpoint)\n tf.train.init_from_checkpoint(init_checkpoint, assignment_map)\n \n output_spec = None\n if mode == tf.estimator.ModeKeys.TRAIN:\n train_op = optimization.create_optimizer(\n total_loss, learning_rate, num_train_steps, num_warmup_steps, False)\n hook_dict = {}\n hook_dict['loss'] = total_loss\n hook_dict['global_steps'] = tf.train.get_or_create_global_step()\n logging_hook = tf.train.LoggingTensorHook(\n hook_dict, every_n_iter=100)\n\n output_spec = tf.estimator.EstimatorSpec(\n mode=mode,\n loss=total_loss,\n train_op=train_op,\n training_hooks=[logging_hook])\n\n elif mode == tf.estimator.ModeKeys.EVAL:\n def metric_fn(label_ids, pred_ids):\n\n return {\n \"eval_loss\": tf.metrics.mean_squared_error(labels=label_ids, predictions=pred_ids), }\n \n eval_metrics = metric_fn(label_ids, pred_ids)\n output_spec = tf.estimator.EstimatorSpec(\n mode=mode,\n loss=total_loss,\n eval_metric_ops=eval_metrics\n )\n else:\n output_spec = tf.estimator.EstimatorSpec(\n mode=mode,\n predictions=pred_ids\n )\n return output_spec\n\n return model_fn\n", "_____no_output_____" ] ], [ [ "#### 创建模型，开始训练", "_____no_output_____" ] ], [ [ "model_fn = model_fn_builder(\n bert_config=bert_config,\n num_labels=len(label_list) + 1,\n init_checkpoint=init_checkpoint,\n learning_rate=learning_rate,\n num_train_steps=num_train_steps,\n num_warmup_steps=num_warmup_steps,\n use_one_hot_embeddings=False)\n\ndef file_based_input_fn_builder(input_file, seq_length, is_training, drop_remainder):\n name_to_features = {\n \"input_ids\": tf.FixedLenFeature([seq_length], tf.int64),\n \"input_mask\": tf.FixedLenFeature([seq_length], tf.int64),\n \"segment_ids\": tf.FixedLenFeature([seq_length], tf.int64),\n \"label_ids\": tf.FixedLenFeature([seq_length], tf.int64),\n }\n\n def _decode_record(record, name_to_features):\n example = tf.parse_single_example(record, name_to_features)\n for name in list(example.keys()):\n t = example[name]\n if t.dtype == tf.int64:\n t = tf.to_int32(t)\n example[name] = t\n return example\n\n def input_fn(params):\n params[\"batch_size\"] = 32\n batch_size = params[\"batch_size\"]\n d = tf.data.TFRecordDataset(input_file)\n if is_training:\n d = d.repeat()\n d = d.shuffle(buffer_size=300)\n d = d.apply(tf.contrib.data.map_and_batch(\n lambda record: _decode_record(record, name_to_features),\n batch_size=batch_size,\n drop_remainder=drop_remainder\n ))\n return d\n\n return input_fn\n\n#训练输入\ntrain_input_fn = file_based_input_fn_builder(\n input_file=train_file,\n seq_length=max_seq_length,\n is_training=True,\n drop_remainder=True)\n\nnum_train_size = len(train_examples)\n\ntf.logging.info(\"***** Running training *****\")\ntf.logging.info(\" Num examples = %d\", num_train_size)\ntf.logging.info(\" Batch size = %d\", batch_size)\ntf.logging.info(\" Num steps = %d\", num_train_steps)\n\n#模型预测estimator\nestimator = tf.estimator.Estimator(\n model_fn=model_fn,\n config=run_config,\n params={\n 'batch_size':batch_size\n })\n\nestimator.train(input_fn=train_input_fn, max_steps=num_train_steps)\n", "INFO:tensorflow:***** Running training *****\nINFO:tensorflow: Num examples = 20864\nINFO:tensorflow: Batch size = 64\nINFO:tensorflow: Num steps = 1630\nINFO:tensorflow:Using config: {'_model_dir': './ner/output', '_tf_random_seed': None, '_save_summary_steps': 1000, '_save_checkpoints_steps': 1000, '_save_checkpoints_secs': None, '_session_config': allow_soft_placement: true\n, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_service': None, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x7fca68ba6748>, '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}\nINFO:tensorflow:Calling model_fn.\nINFO:tensorflow:*** Features ***\nINFO:tensorflow: name = input_ids, shape = (32, 128)\nINFO:tensorflow: name = input_mask, shape = (32, 128)\nINFO:tensorflow: name = label_ids, shape = (32, 128)\nINFO:tensorflow: name = segment_ids, shape = (32, 128)\n" ] ], [ [ "#### 在验证集上验证模型", "_____no_output_____" ] ], [ [ "eval_examples = processor.get_dev_examples(data_dir)\neval_file = os.path.join(output_dir, \"eval.tf_record\")\nfiled_based_convert_examples_to_features(\n eval_examples, label_list, max_seq_length, tokenizer, eval_file)\ndata_config['eval.tf_record_path'] = eval_file\ndata_config['num_eval_size'] = len(eval_examples)\nnum_eval_size = data_config.get('num_eval_size', 0)\n\ntf.logging.info(\"***** Running evaluation *****\")\ntf.logging.info(\" Num examples = %d\", num_eval_size)\ntf.logging.info(\" Batch size = %d\", batch_size)\n\neval_steps = None\neval_drop_remainder = False\neval_input_fn = file_based_input_fn_builder(\n input_file=eval_file,\n seq_length=max_seq_length,\n is_training=False,\n drop_remainder=eval_drop_remainder)\n\nresult = estimator.evaluate(input_fn=eval_input_fn, steps=eval_steps)\noutput_eval_file = os.path.join(output_dir, \"eval_results.txt\")\nwith codecs.open(output_eval_file, \"w\", encoding='utf-8') as writer:\n tf.logging.info(\"***** Eval results *****\")\n for key in sorted(result.keys()):\n tf.logging.info(\" %s = %s\", key, str(result[key]))\n writer.write(\"%s = %s\\n\" % (key, str(result[key])))\n\nif not os.path.exists(data_config_path):\n with codecs.open(data_config_path, 'a', encoding='utf-8') as fd:\n json.dump(data_config, fd)\n", "INFO:tensorflow:Writing example 0 of 4631\nINFO:tensorflow:***** Running evaluation *****\nINFO:tensorflow: Num examples = 4631\nINFO:tensorflow: Batch size = 64\nINFO:tensorflow:Calling model_fn.\nINFO:tensorflow:*** Features ***\nINFO:tensorflow: name = input_ids, shape = (?, 128)\nINFO:tensorflow: name = input_mask, shape = (?, 128)\nINFO:tensorflow: name = label_ids, shape = (?, 128)\nINFO:tensorflow: name = segment_ids, shape = (?, 128)\n" ] ], [ [ "#### 在测试集上进行测试", "_____no_output_____" ] ], [ [ "token_path = os.path.join(output_dir, \"token_test.txt\")\nif os.path.exists(token_path):\n os.remove(token_path)\n\nwith codecs.open(os.path.join(output_dir, 'label2id.pkl'), 'rb') as rf:\n label2id = pickle.load(rf)\n id2label = {value: key for key, value in label2id.items()}\n\npredict_examples = processor.get_test_examples(data_dir)\npredict_file = os.path.join(output_dir, \"predict.tf_record\")\nfiled_based_convert_examples_to_features(predict_examples, label_list,\n max_seq_length, tokenizer,\n predict_file, mode=\"test\")\n\ntf.logging.info(\"***** Running prediction*****\")\ntf.logging.info(\" Num examples = %d\", len(predict_examples))\ntf.logging.info(\" Batch size = %d\", batch_size)\n \npredict_drop_remainder = False\npredict_input_fn = file_based_input_fn_builder(\n input_file=predict_file,\n seq_length=max_seq_length,\n is_training=False,\n drop_remainder=predict_drop_remainder)\n\npredicted_result = estimator.evaluate(input_fn=predict_input_fn)\noutput_eval_file = os.path.join(output_dir, \"predicted_results.txt\")\nwith codecs.open(output_eval_file, \"w\", encoding='utf-8') as writer:\n tf.logging.info(\"***** Predict results *****\")\n for key in sorted(predicted_result.keys()):\n tf.logging.info(\" %s = %s\", key, str(predicted_result[key]))\n writer.write(\"%s = %s\\n\" % (key, str(predicted_result[key])))\n\nresult = estimator.predict(input_fn=predict_input_fn)\noutput_predict_file = os.path.join(output_dir, \"label_test.txt\")\n\ndef result_to_pair(writer):\n for predict_line, prediction in zip(predict_examples, result):\n idx = 0\n line = ''\n line_token = str(predict_line.text).split(' ')\n label_token = str(predict_line.label).split(' ')\n if len(line_token) != len(label_token):\n tf.logging.info(predict_line.text)\n tf.logging.info(predict_line.label)\n for id in prediction:\n if id == 0:\n continue\n curr_labels = id2label[id]\n if curr_labels in ['[CLS]', '[SEP]']:\n continue\n try:\n line += line_token[idx] + ' ' + label_token[idx] + ' ' + curr_labels + '\\n'\n except Exception as e:\n tf.logging.info(e)\n tf.logging.info(predict_line.text)\n tf.logging.info(predict_line.label)\n line = ''\n break\n idx += 1\n writer.write(line + '\\n')\n \nfrom ner.src.conlleval import return_report\n\nwith codecs.open(output_predict_file, 'w', encoding='utf-8') as writer:\n result_to_pair(writer)\neval_result = return_report(output_predict_file)\nfor line in eval_result:\n print(line)", "INFO:tensorflow:Writing example 0 of 68\nINFO:tensorflow:***** Running prediction*****\nINFO:tensorflow: Num examples = 68\nINFO:tensorflow: Batch size = 64\nINFO:tensorflow:Calling model_fn.\nINFO:tensorflow:*** Features ***\nINFO:tensorflow: name = input_ids, shape = (?, 128)\nINFO:tensorflow: name = input_mask, shape = (?, 128)\nINFO:tensorflow: name = label_ids, shape = (?, 128)\nINFO:tensorflow: name = segment_ids, shape = (?, 128)\n" ] ], [ [ "### 在线命名实体识别\n\n由以上训练得到模型进行在线测试，可以任意输入句子，进行命名实体识别。\n\n输入“再见”，结束在线命名实体识别。\n\n<span style=\"color:red\">若下述程序未执行成功，则表示训练完成后，GPU显存还在占用，需要restart kernel，然后执行 %run 命令。</span>\n\n释放资源具体流程为：菜单 > Kernel > Restart \n\n\n", "_____no_output_____" ] ], [ [ "%run ner/src/terminal_predict.py", "checkpoint path:./ner/output/checkpoint\ngoing to restore checkpoint\nINFO:tensorflow:Restoring parameters from ./ner/output/model.ckpt-1630\n{1: 'O', 2: 'B-PER', 3: 'I-PER', 4: 'B-ORG', 5: 'I-ORG', 6: 'B-LOC', 7: 'I-LOC', 8: 'X', 9: '[CLS]', 10: '[SEP]'}\n输入句子:\n中国男篮与委内瑞拉队在北京五棵松体育馆展开小组赛最后一场比赛的争夺，赵继伟12分4助攻3抢断、易建联11分8篮板、周琦8分7篮板2盖帽。\n[['B-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'O', 'B-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'O', 'B-LOC', 'I-LOC', 'B-LOC', 'I-LOC', 'I-LOC', 'I-LOC', 'I-LOC', 'I-LOC', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-PER', 'I-PER', 'I-PER', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-PER', 'I-PER', 'I-PER', 'O', 'O', 'O', 'O', 'O', 'O', 'B-PER', 'I-PER', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O']]\nLOC, 北京, 五棵松体育馆\nPER, 赵继伟, 易建联, 周琦\nORG, 中国男篮, 委内瑞拉队\ntime used: 0.908481 sec\n输入句子:\n周杰伦（Jay Chou），1979年1月18日出生于台湾省新北市，毕业于淡江中学，中国台湾流行乐男歌手。\n[['B-PER', 'I-PER', 'I-PER', 'O', 'B-PER', 'I-PER', 'I-PER', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-LOC', 'I-LOC', 'I-LOC', 'B-LOC', 'I-LOC', 'I-LOC', 'O', 'O', 'O', 'O', 'B-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'O', 'B-LOC', 'I-LOC', 'B-LOC', 'I-LOC', 'O', 'O', 'O', 'O', 'O', 'O', 'O']]\nLOC, 台湾省, 新北市, 中国, 台湾\nPER, 周杰伦, jaycho##u\nORG, 淡江中学\ntime used: 0.058148 sec\n输入句子:\n马云，1964年9月10日生于浙江省杭州市，1988年毕业于杭州师范学院外语系，同年担任杭州电子工业学院英文及国际贸易教师。\n[['B-PER', 'I-PER', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'B-LOC', 'I-LOC', 'I-LOC', 'B-LOC', 'I-LOC', 'I-LOC', 'O', 'O', 'O', 'O', 'O', 'O', 'B-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'O', 'O', 'O', 'O', 'O', 'B-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'I-ORG', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O', 'O']]\nLOC, 浙江省, 杭州市\nPER, 马云\nORG, 杭州师范学院外语系, 杭州电子工业学院\ntime used: 0.065471 sec\n输入句子:\n再见\n\n再见\n" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# 100 pandas puzzles\n\nInspired by [100 Numpy exerises](https://github.com/rougier/numpy-100), here are 100* short puzzles for testing your knowledge of [pandas'](http://pandas.pydata.org/) power.\n\nSince pandas is a large library with many different specialist features and functions, these excercises focus mainly on the fundamentals of manipulating data (indexing, grouping, aggregating, cleaning), making use of the core DataFrame and Series objects. \n\nMany of the excerises here are stright-forward in that the solutions require no more than a few lines of code (in pandas or NumPy... don't go using pure Python or Cython!). Choosing the right methods and following best practices is the underlying goal.\n\nThe exercises are loosely divided in sections. Each section has a difficulty rating; these ratings are subjective, of course, but should be a seen as a rough guide as to how inventive the required solution is.\n\nIf you're just starting out with pandas and you are looking for some other resources, the official documentation is very extensive. In particular, some good places get a broader overview of pandas are...\n\n- [10 minutes to pandas](http://pandas.pydata.org/pandas-docs/stable/10min.html)\n- [pandas basics](http://pandas.pydata.org/pandas-docs/stable/basics.html)\n- [tutorials](http://pandas.pydata.org/pandas-docs/stable/tutorials.html)\n- [cookbook and idioms](http://pandas.pydata.org/pandas-docs/stable/cookbook.html#cookbook)\n\nEnjoy the puzzles!\n\n\\* *the list of exercises is not yet complete! Pull requests or suggestions for additional exercises, corrections and improvements are welcomed.*", "_____no_output_____" ], [ "## Importing pandas\n\n### Getting started and checking your pandas setup\n\nDifficulty: *easy* \n\n**1.** Import pandas under the alias `pd`.", "_____no_output_____" ] ], [ [ "import pandas as pd", "_____no_output_____" ] ], [ [ "**2.** Print the version of pandas that has been imported.", "_____no_output_____" ] ], [ [ "pd.__version__", "_____no_output_____" ] ], [ [ "**3.** Print out all the *version* information of the libraries that are required by the pandas library.", "_____no_output_____" ] ], [ [ "pd.show_versions()", "\nINSTALLED VERSIONS\n------------------\ncommit : 2cb96529396d93b46abab7bbc73a208e708c642e\npython : 3.8.8.final.0\npython-bits : 64\nOS : Windows\nOS-release : 10\nVersion : 10.0.22000\nmachine : AMD64\nprocessor : Intel64 Family 6 Model 142 Stepping 10, GenuineIntel\nbyteorder : little\nLC_ALL : None\nLANG : None\nLOCALE : English_United States.1252\n\npandas : 1.2.4\nnumpy : 1.20.1\npytz : 2020.1\ndateutil : 2.8.1\npip : 20.2.2\nsetuptools : 52.0.0.post20210125\nCython : 0.29.23\npytest : 6.2.3\nhypothesis : None\nsphinx : 4.0.1\nblosc : None\nfeather : None\nxlsxwriter : 1.3.8\nlxml.etree : 4.6.3\nhtml5lib : 1.1\npymysql : None\npsycopg2 : 2.8.6 (dt dec pq3 ext lo64)\njinja2 : 2.11.3\nIPython : 7.22.0\npandas_datareader: 0.10.0\nbs4 : 4.9.3\nbottleneck : 1.3.2\nfsspec : 0.9.0\nfastparquet : None\ngcsfs : None\nmatplotlib : 3.3.4\nnumexpr : 2.7.3\nodfpy : None\nopenpyxl : 3.0.7\npandas_gbq : None\npyarrow : None\npyxlsb : None\ns3fs : None\nscipy : 1.6.2\nsqlalchemy : 1.4.7\ntables : 3.6.1\ntabulate : None\nxarray : None\nxlrd : 2.0.1\nxlwt : 1.3.0\nnumba : 0.53.1\n" ] ], [ [ "## DataFrame basics\n\n### A few of the fundamental routines for selecting, sorting, adding and aggregating data in DataFrames\n\nDifficulty: *easy*\n\nNote: remember to import numpy using:\n```python\nimport numpy as np\n```\n\nConsider the following Python dictionary `data` and Python list `labels`:\n\n``` python\ndata = {'animal': ['cat', 'cat', 'snake', 'dog', 'dog', 'cat', 'snake', 'cat', 'dog', 'dog'],\n 'age': [2.5, 3, 0.5, np.nan, 5, 2, 4.5, np.nan, 7, 3],\n 'visits': [1, 3, 2, 3, 2, 3, 1, 1, 2, 1],\n 'priority': ['yes', 'yes', 'no', 'yes', 'no', 'no', 'no', 'yes', 'no', 'no']}\n\nlabels = ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j']\n```\n(This is just some meaningless data I made up with the theme of animals and trips to a vet.)\n\n**4.** Create a DataFrame `df` from this dictionary `data` which has the index `labels`.", "_____no_output_____" ] ], [ [ "import numpy as np\n\nraw_data = {'animal': ['cat', 'cat', 'snake', 'dog', 'dog', 'cat', 'snake', 'cat', 'dog', 'dog'],\n 'age': [2.5, 3, 0.5, np.nan, 5, 2, 4.5, np.nan, 7, 3],\n 'visits': [1, 3, 2, 3, 2, 3, 1, 1, 2, 1],\n 'priority': ['yes', 'yes', 'no', 'yes', 'no', 'no', 'no', 'yes', 'no', 'no']}\n\nlabels = ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j']\n\ndf = pd.DataFrame(raw_data, index=labels)# (complete this line of code)\ndf", "_____no_output_____" ] ], [ [ "**5.** Display a summary of the basic information about this DataFrame and its data (*hint: there is a single method that can be called on the DataFrame*).", "_____no_output_____" ] ], [ [ "df.describe()", "_____no_output_____" ] ], [ [ "**6.** Return the first 3 rows of the DataFrame `df`.", "_____no_output_____" ] ], [ [ "df.iloc[:3,:]", "_____no_output_____" ] ], [ [ "**7.** Select just the 'animal' and 'age' columns from the DataFrame `df`.", "_____no_output_____" ] ], [ [ "df[['animal', 'age']]", "_____no_output_____" ] ], [ [ "**8.** Select the data in rows `[3, 4, 8]` *and* in columns `['animal', 'age']`.", "_____no_output_____" ] ], [ [ "df.iloc[[3, 4, 8]][['animal','age']]", "_____no_output_____" ] ], [ [ "**9.** Select only the rows where the number of visits is greater than 3.", "_____no_output_____" ] ], [ [ "df[df['visits'] > 3]", "_____no_output_____" ] ], [ [ "**10.** Select the rows where the age is missing, i.e. it is `NaN`.", "_____no_output_____" ] ], [ [ "df[df['age'].isna()]", "_____no_output_____" ] ], [ [ "**11.** Select the rows where the animal is a cat *and* the age is less than 3.", "_____no_output_____" ] ], [ [ "df[(df['animal'] == 'cat') & (df['age'] < 3)]", "_____no_output_____" ] ], [ [ "**12.** Select the rows the age is between 2 and 4 (inclusive).", "_____no_output_____" ] ], [ [ "df.iloc[2:5]", "_____no_output_____" ] ], [ [ "**13.** Change the age in row 'f' to 1.5.", "_____no_output_____" ] ], [ [ "df.loc['f','age'] = 1.5\ndf", "_____no_output_____" ] ], [ [ "**14.** Calculate the sum of all visits in `df` (i.e. find the total number of visits).", "_____no_output_____" ] ], [ [ "df['visits'].sum()", "_____no_output_____" ] ], [ [ "**15.** Calculate the mean age for each different animal in `df`.", "_____no_output_____" ] ], [ [ "df.groupby('animal').agg({'age':'mean'})", "_____no_output_____" ] ], [ [ "**16.** Append a new row 'k' to `df` with your choice of values for each column. Then delete that row to return the original DataFrame.", "_____no_output_____" ] ], [ [ "import numpy as np\nrnum = np.random.randint(10, size=len(df))\ndf['k'] = rnum", "_____no_output_____" ] ], [ [ "**17.** Count the number of each type of animal in `df`.", "_____no_output_____" ] ], [ [ "df['animal'].value_counts()", "_____no_output_____" ] ], [ [ "**18.** Sort `df` first by the values in the 'age' in *decending* order, then by the value in the 'visits' column in *ascending* order (so row `i` should be first, and row `d` should be last).", "_____no_output_____" ] ], [ [ "df.sort_values(by = ['age', 'visits'], ascending=[False, True])", "_____no_output_____" ] ], [ [ "**19.** The 'priority' column contains the values 'yes' and 'no'. Replace this column with a column of boolean values: 'yes' should be `True` and 'no' should be `False`.", "_____no_output_____" ] ], [ [ "df['priority'] = df['priority'].map({'yes': True, 'no': False})", "_____no_output_____" ] ], [ [ "**20.** In the 'animal' column, change the 'snake' entries to 'python'.", "_____no_output_____" ] ], [ [ "df['animal'] = df['animal'].replace('snake', 'python')\ndf", "_____no_output_____" ] ], [ [ "**21.** For each animal type and each number of visits, find the mean age. In other words, each row is an animal, each column is a number of visits and the values are the mean ages (*hint: use a pivot table*).", "_____no_output_____" ] ], [ [ "df.pivot_table(index = 'animal', columns = 'visits', values = 'age', aggfunc = 'mean').fillna(0)", "_____no_output_____" ] ], [ [ "## DataFrames: beyond the basics\n\n### Slightly trickier: you may need to combine two or more methods to get the right answer\n\nDifficulty: *medium*\n\nThe previous section was tour through some basic but essential DataFrame operations. Below are some ways that you might need to cut your data, but for which there is no single \"out of the box\" method.", "_____no_output_____" ], [ "**22.** You have a DataFrame `df` with a column 'A' of integers. For example:\n```python\ndf = pd.DataFrame({'A': [1, 2, 2, 3, 4, 5, 5, 5, 6, 7, 7]})\n```\n\nHow do you filter out rows which contain the same integer as the row immediately above?\n\nYou should be left with a column containing the following values:\n\n```python\n1, 2, 3, 4, 5, 6, 7\n```", "_____no_output_____" ], [ "**23.** Given a DataFrame of numeric values, say\n```python\ndf = pd.DataFrame(np.random.random(size=(5, 3))) # a 5x3 frame of float values\n```\n\nhow do you subtract the row mean from each element in the row?", "_____no_output_____" ], [ "**24.** Suppose you have DataFrame with 10 columns of real numbers, for example:\n\n```python\ndf = pd.DataFrame(np.random.random(size=(5, 10)), columns=list('abcdefghij'))\n```\nWhich column of numbers has the smallest sum? Return that column's label.", "_____no_output_____" ], [ "**25.** How do you count how many unique rows a DataFrame has (i.e. ignore all rows that are duplicates)? As input, use a DataFrame of zeros and ones with 10 rows and 3 columns.\n\n```python\ndf = pd.DataFrame(np.random.randint(0, 2, size=(10, 3)))\n```", "_____no_output_____" ], [ "The next three puzzles are slightly harder.\n\n\n**26.** In the cell below, you have a DataFrame `df` that consists of 10 columns of floating-point numbers. Exactly 5 entries in each row are NaN values. \n\nFor each row of the DataFrame, find the *column* which contains the *third* NaN value.\n\nYou should return a Series of column labels: `e, c, d, h, d`", "_____no_output_____" ] ], [ [ "nan = np.nan\n\ndata = [[0.04, nan, nan, 0.25, nan, 0.43, 0.71, 0.51, nan, nan],\n [ nan, nan, nan, 0.04, 0.76, nan, nan, 0.67, 0.76, 0.16],\n [ nan, nan, 0.5 , nan, 0.31, 0.4 , nan, nan, 0.24, 0.01],\n [0.49, nan, nan, 0.62, 0.73, 0.26, 0.85, nan, nan, nan],\n [ nan, nan, 0.41, nan, 0.05, nan, 0.61, nan, 0.48, 0.68]]\n\ncolumns = list('abcdefghij')\n\ndf = pd.DataFrame(data, columns=columns)\n\n# write a solution to the question here", "_____no_output_____" ] ], [ [ "**27.** A DataFrame has a column of groups 'grps' and and column of integer values 'vals': \n\n```python\ndf = pd.DataFrame({'grps': list('aaabbcaabcccbbc'), \n 'vals': [12,345,3,1,45,14,4,52,54,23,235,21,57,3,87]})\n```\nFor each *group*, find the sum of the three greatest values. You should end up with the answer as follows:\n```\ngrps\na 409\nb 156\nc 345\n```", "_____no_output_____" ] ], [ [ "df = pd.DataFrame({'grps': list('aaabbcaabcccbbc'), \n 'vals': [12,345,3,1,45,14,4,52,54,23,235,21,57,3,87]})\n\n# write a solution to the question here", "_____no_output_____" ] ], [ [ "**28.** The DataFrame `df` constructed below has two integer columns 'A' and 'B'. The values in 'A' are between 1 and 100 (inclusive). \n\nFor each group of 10 consecutive integers in 'A' (i.e. `(0, 10]`, `(10, 20]`, ...), calculate the sum of the corresponding values in column 'B'.\n\nThe answer should be a Series as follows:\n\n```\nA\n(0, 10] 635\n(10, 20] 360\n(20, 30] 315\n(30, 40] 306\n(40, 50] 750\n(50, 60] 284\n(60, 70] 424\n(70, 80] 526\n(80, 90] 835\n(90, 100] 852\n```", "_____no_output_____" ] ], [ [ "df = pd.DataFrame(np.random.RandomState(8765).randint(1, 101, size=(100, 2)), columns = [\"A\", \"B\"])\n\n# write a solution to the question here", "_____no_output_____" ] ], [ [ "## DataFrames: harder problems \n\n### These might require a bit of thinking outside the box...\n\n...but all are solvable using just the usual pandas/NumPy methods (and so avoid using explicit `for` loops).\n\nDifficulty: *hard*", "_____no_output_____" ], [ "**29.** Consider a DataFrame `df` where there is an integer column 'X':\n```python\ndf = pd.DataFrame({'X': [7, 2, 0, 3, 4, 2, 5, 0, 3, 4]})\n```\nFor each value, count the difference back to the previous zero (or the start of the Series, whichever is closer). These values should therefore be \n\n```\n[1, 2, 0, 1, 2, 3, 4, 0, 1, 2]\n```\n\nMake this a new column 'Y'.", "_____no_output_____" ], [ "**30.** Consider the DataFrame constructed below which contains rows and columns of numerical data. \n\nCreate a list of the column-row index locations of the 3 largest values in this DataFrame. In this case, the answer should be:\n```\n[(5, 7), (6, 4), (2, 5)]\n```", "_____no_output_____" ] ], [ [ "df = pd.DataFrame(np.random.RandomState(30).randint(1, 101, size=(8, 8)))", "_____no_output_____" ] ], [ [ "**31.** You are given the DataFrame below with a column of group IDs, 'grps', and a column of corresponding integer values, 'vals'.\n\n```python\ndf = pd.DataFrame({\"vals\": np.random.RandomState(31).randint(-30, 30, size=15), \n \"grps\": np.random.RandomState(31).choice([\"A\", \"B\"], 15)})\n```\n\nCreate a new column 'patched_values' which contains the same values as the 'vals' any negative values in 'vals' with the group mean:\n\n```\n vals grps patched_vals\n0 -12 A 13.6\n1 -7 B 28.0\n2 -14 A 13.6\n3 4 A 4.0\n4 -7 A 13.6\n5 28 B 28.0\n6 -2 A 13.6\n7 -1 A 13.6\n8 8 A 8.0\n9 -2 B 28.0\n10 28 A 28.0\n11 12 A 12.0\n12 16 A 16.0\n13 -24 A 13.6\n14 -12 A 13.6\n```", "_____no_output_____" ], [ "**32.** Implement a rolling mean over groups with window size 3, which ignores NaN value. For example consider the following DataFrame:\n\n```python\n>>> df = pd.DataFrame({'group': list('aabbabbbabab'),\n 'value': [1, 2, 3, np.nan, 2, 3, np.nan, 1, 7, 3, np.nan, 8]})\n>>> df\n group value\n0 a 1.0\n1 a 2.0\n2 b 3.0\n3 b NaN\n4 a 2.0\n5 b 3.0\n6 b NaN\n7 b 1.0\n8 a 7.0\n9 b 3.0\n10 a NaN\n11 b 8.0\n```\nThe goal is to compute the Series:\n\n```\n0 1.000000\n1 1.500000\n2 3.000000\n3 3.000000\n4 1.666667\n5 3.000000\n6 3.000000\n7 2.000000\n8 3.666667\n9 2.000000\n10 4.500000\n11 4.000000\n```\nE.g. the first window of size three for group 'b' has values 3.0, NaN and 3.0 and occurs at row index 5. Instead of being NaN the value in the new column at this row index should be 3.0 (just the two non-NaN values are used to compute the mean (3+3)/2)", "_____no_output_____" ], [ "## Series and DatetimeIndex\n\n### Exercises for creating and manipulating Series with datetime data\n\nDifficulty: *easy/medium*\n\npandas is fantastic for working with dates and times. These puzzles explore some of this functionality.\n", "_____no_output_____" ], [ "**33.** Create a DatetimeIndex that contains each business day of 2015 and use it to index a Series of random numbers. Let's call this Series `s`.", "_____no_output_____" ], [ "**34.** Find the sum of the values in `s` for every Wednesday.", "_____no_output_____" ], [ "**35.** For each calendar month in `s`, find the mean of values.", "_____no_output_____" ], [ "**36.** For each group of four consecutive calendar months in `s`, find the date on which the highest value occurred.", "_____no_output_____" ], [ "**37.** Create a DateTimeIndex consisting of the third Thursday in each month for the years 2015 and 2016.", "_____no_output_____" ], [ "## Cleaning Data\n\n### Making a DataFrame easier to work with\n\nDifficulty: *easy/medium*\n\nIt happens all the time: someone gives you data containing malformed strings, Python, lists and missing data. How do you tidy it up so you can get on with the analysis?\n\nTake this monstrosity as the DataFrame to use in the following puzzles:\n\n```python\ndf = pd.DataFrame({'From_To': ['LoNDon_paris', 'MAdrid_miLAN', 'londON_StockhOlm', \n 'Budapest_PaRis', 'Brussels_londOn'],\n 'FlightNumber': [10045, np.nan, 10065, np.nan, 10085],\n 'RecentDelays': [[23, 47], [], [24, 43, 87], [13], [67, 32]],\n 'Airline': ['KLM(!)', '<Air France> (12)', '(British Airways. )', \n '12. Air France', '\"Swiss Air\"']})\n```\nFormatted, it looks like this:\n\n```\n From_To FlightNumber RecentDelays Airline\n0 LoNDon_paris 10045.0 [23, 47] KLM(!)\n1 MAdrid_miLAN NaN [] <Air France> (12)\n2 londON_StockhOlm 10065.0 [24, 43, 87] (British Airways. )\n3 Budapest_PaRis NaN [13] 12. Air France\n4 Brussels_londOn 10085.0 [67, 32] \"Swiss Air\"\n```\n\n\n(It's some flight data I made up; it's not meant to be accurate in any way.)\n", "_____no_output_____" ], [ "**38.** Some values in the the **FlightNumber** column are missing (they are `NaN`). These numbers are meant to increase by 10 with each row so 10055 and 10075 need to be put in place. Modify `df` to fill in these missing numbers and make the column an integer column (instead of a float column).", "_____no_output_____" ], [ "**39.** The **From\\_To** column would be better as two separate columns! Split each string on the underscore delimiter `_` to give a new temporary DataFrame called 'temp' with the correct values. Assign the correct column names 'From' and 'To' to this temporary DataFrame. ", "_____no_output_____" ], [ "**40.** Notice how the capitalisation of the city names is all mixed up in this temporary DataFrame 'temp'. Standardise the strings so that only the first letter is uppercase (e.g. \"londON\" should become \"London\".)", "_____no_output_____" ], [ "**41.** Delete the **From_To** column from `df` and attach the temporary DataFrame 'temp' from the previous questions.", "_____no_output_____" ], [ "**42**. In the **Airline** column, you can see some extra puctuation and symbols have appeared around the airline names. Pull out just the airline name. E.g. `'(British Airways. )'` should become `'British Airways'`.", "_____no_output_____" ], [ "**43**. In the RecentDelays column, the values have been entered into the DataFrame as a list. We would like each first value in its own column, each second value in its own column, and so on. If there isn't an Nth value, the value should be NaN.\n\nExpand the Series of lists into a DataFrame named `delays`, rename the columns `delay_1`, `delay_2`, etc. and replace the unwanted RecentDelays column in `df` with `delays`.", "_____no_output_____" ], [ "The DataFrame should look much better now.\n```\n FlightNumber Airline From To delay_1 delay_2 delay_3\n0 10045 KLM London Paris 23.0 47.0 NaN\n1 10055 Air France Madrid Milan NaN NaN NaN\n2 10065 British Airways London Stockholm 24.0 43.0 87.0\n3 10075 Air France Budapest Paris 13.0 NaN NaN\n4 10085 Swiss Air Brussels London 67.0 32.0 NaN\n```", "_____no_output_____" ], [ "## Using MultiIndexes\n\n### Go beyond flat DataFrames with additional index levels\n\nDifficulty: *medium*\n\nPrevious exercises have seen us analysing data from DataFrames equipped with a single index level. However, pandas also gives you the possibilty of indexing your data using *multiple* levels. This is very much like adding new dimensions to a Series or a DataFrame. For example, a Series is 1D, but by using a MultiIndex with 2 levels we gain of much the same functionality as a 2D DataFrame.\n\nThe set of puzzles below explores how you might use multiple index levels to enhance data analysis.\n\nTo warm up, we'll look make a Series with two index levels. ", "_____no_output_____" ], [ "**44**. Given the lists `letters = ['A', 'B', 'C']` and `numbers = list(range(10))`, construct a MultiIndex object from the product of the two lists. Use it to index a Series of random numbers. Call this Series `s`.", "_____no_output_____" ], [ "**45.** Check the index of `s` is lexicographically sorted (this is a necessary proprty for indexing to work correctly with a MultiIndex).", "_____no_output_____" ], [ "**46**. Select the labels `1`, `3` and `6` from the second level of the MultiIndexed Series.", "_____no_output_____" ], [ "**47**. Slice the Series `s`; slice up to label 'B' for the first level and from label 5 onwards for the second level.", "_____no_output_____" ], [ "**48**. Sum the values in `s` for each label in the first level (you should have Series giving you a total for labels A, B and C).", "_____no_output_____" ], [ "**49**. Suppose that `sum()` (and other methods) did not accept a `level` keyword argument. How else could you perform the equivalent of `s.sum(level=1)`?", "_____no_output_____" ], [ "**50**. Exchange the levels of the MultiIndex so we have an index of the form (letters, numbers). Is this new Series properly lexsorted? If not, sort it.", "_____no_output_____" ], [ "## Minesweeper\n\n### Generate the numbers for safe squares in a Minesweeper grid\n\nDifficulty: *medium* to *hard*\n\nIf you've ever used an older version of Windows, there's a good chance you've played with Minesweeper:\n- https://en.wikipedia.org/wiki/Minesweeper_(video_game)\n\n\nIf you're not familiar with the game, imagine a grid of squares: some of these squares conceal a mine. If you click on a mine, you lose instantly. If you click on a safe square, you reveal a number telling you how many mines are found in the squares that are immediately adjacent. The aim of the game is to uncover all squares in the grid that do not contain a mine.\n\nIn this section, we'll make a DataFrame that contains the necessary data for a game of Minesweeper: coordinates of the squares, whether the square contains a mine and the number of mines found on adjacent squares.", "_____no_output_____" ], [ "**51**. Let's suppose we're playing Minesweeper on a 5 by 4 grid, i.e.\n```\nX = 5\nY = 4\n```\nTo begin, generate a DataFrame `df` with two columns, `'x'` and `'y'` containing every coordinate for this grid. That is, the DataFrame should start:\n```\n x y\n0 0 0\n1 0 1\n2 0 2\n```", "_____no_output_____" ], [ "**52**. For this DataFrame `df`, create a new column of zeros (safe) and ones (mine). The probability of a mine occuring at each location should be 0.4.", "_____no_output_____" ], [ "**53**. Now create a new column for this DataFrame called `'adjacent'`. This column should contain the number of mines found on adjacent squares in the grid. \n\n(E.g. for the first row, which is the entry for the coordinate `(0, 0)`, count how many mines are found on the coordinates `(0, 1)`, `(1, 0)` and `(1, 1)`.)", "_____no_output_____" ], [ "**54**. For rows of the DataFrame that contain a mine, set the value in the `'adjacent'` column to NaN.", "_____no_output_____" ], [ "**55**. Finally, convert the DataFrame to grid of the adjacent mine counts: columns are the `x` coordinate, rows are the `y` coordinate.", "_____no_output_____" ], [ "## Plotting\n\n### Visualize trends and patterns in data\n\nDifficulty: *medium*\n\nTo really get a good understanding of the data contained in your DataFrame, it is often essential to create plots: if you're lucky, trends and anomalies will jump right out at you. This functionality is baked into pandas and the puzzles below explore some of what's possible with the library.\n\n**56.** Pandas is highly integrated with the plotting library matplotlib, and makes plotting DataFrames very user-friendly! Plotting in a notebook environment usually makes use of the following boilerplate:\n\n```python\nimport matplotlib.pyplot as plt\n%matplotlib inline\nplt.style.use('ggplot')\n```\n\nmatplotlib is the plotting library which pandas' plotting functionality is built upon, and it is usually aliased to ```plt```.\n\n```%matplotlib inline``` tells the notebook to show plots inline, instead of creating them in a separate window. \n\n```plt.style.use('ggplot')``` is a style theme that most people find agreeable, based upon the styling of R's ggplot package.\n\nFor starters, make a scatter plot of this random data, but use black X's instead of the default markers. \n\n```df = pd.DataFrame({\"xs\":[1,5,2,8,1], \"ys\":[4,2,1,9,6]})```\n\nConsult the [documentation](https://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.plot.html) if you get stuck!", "_____no_output_____" ], [ "**57.** Columns in your DataFrame can also be used to modify colors and sizes. Bill has been keeping track of his performance at work over time, as well as how good he was feeling that day, and whether he had a cup of coffee in the morning. Make a plot which incorporates all four features of this DataFrame.\n\n(Hint: If you're having trouble seeing the plot, try multiplying the Series which you choose to represent size by 10 or more)\n\n*The chart doesn't have to be pretty: this isn't a course in data viz!*\n\n```\ndf = pd.DataFrame({\"productivity\":[5,2,3,1,4,5,6,7,8,3,4,8,9],\n \"hours_in\" :[1,9,6,5,3,9,2,9,1,7,4,2,2],\n \"happiness\" :[2,1,3,2,3,1,2,3,1,2,2,1,3],\n \"caffienated\" :[0,0,1,1,0,0,0,0,1,1,0,1,0]})\n```", "_____no_output_____" ], [ "**58.** What if we want to plot multiple things? Pandas allows you to pass in a matplotlib *Axis* object for plots, and plots will also return an Axis object.\n\nMake a bar plot of monthly revenue with a line plot of monthly advertising spending (numbers in millions)\n\n```\ndf = pd.DataFrame({\"revenue\":[57,68,63,71,72,90,80,62,59,51,47,52],\n \"advertising\":[2.1,1.9,2.7,3.0,3.6,3.2,2.7,2.4,1.8,1.6,1.3,1.9],\n \"month\":range(12)\n })\n```", "_____no_output_____" ], [ "Now we're finally ready to create a candlestick chart, which is a very common tool used to analyze stock price data. A candlestick chart shows the opening, closing, highest, and lowest price for a stock during a time window. The color of the \"candle\" (the thick part of the bar) is green if the stock closed above its opening price, or red if below.\n\n\n\nThis was initially designed to be a pandas plotting challenge, but it just so happens that this type of plot is just not feasible using pandas' methods. If you are unfamiliar with matplotlib, we have provided a function that will plot the chart for you so long as you can use pandas to get the data into the correct format.\n\nYour first step should be to get the data in the correct format using pandas' time-series grouping function. We would like each candle to represent an hour's worth of data. You can write your own aggregation function which returns the open/high/low/close, but pandas has a built-in which also does this.", "_____no_output_____" ], [ "The below cell contains helper functions. Call ```day_stock_data()``` to generate a DataFrame containing the prices a hypothetical stock sold for, and the time the sale occurred. Call ```plot_candlestick(df)``` on your properly aggregated and formatted stock data to print the candlestick chart.", "_____no_output_____" ] ], [ [ "import numpy as np\ndef float_to_time(x):\n return str(int(x)) + \":\" + str(int(x%1 * 60)).zfill(2) + \":\" + str(int(x*60 % 1 * 60)).zfill(2)\n\ndef day_stock_data():\n #NYSE is open from 9:30 to 4:00\n time = 9.5\n price = 100\n results = [(float_to_time(time), price)]\n while time < 16:\n elapsed = np.random.exponential(.001)\n time += elapsed\n if time > 16:\n break\n price_diff = np.random.uniform(.999, 1.001)\n price *= price_diff\n results.append((float_to_time(time), price))\n \n \n df = pd.DataFrame(results, columns = ['time','price'])\n df.time = pd.to_datetime(df.time)\n return df\n\n#Don't read me unless you get stuck!\ndef plot_candlestick(agg):\n \"\"\"\n agg is a DataFrame which has a DatetimeIndex and five columns: [\"open\",\"high\",\"low\",\"close\",\"color\"]\n \"\"\"\n fig, ax = plt.subplots()\n for time in agg.index:\n ax.plot([time.hour] * 2, agg.loc[time, [\"high\",\"low\"]].values, color = \"black\")\n ax.plot([time.hour] * 2, agg.loc[time, [\"open\",\"close\"]].values, color = agg.loc[time, \"color\"], linewidth = 10)\n\n ax.set_xlim((8,16))\n ax.set_ylabel(\"Price\")\n ax.set_xlabel(\"Hour\")\n ax.set_title(\"OHLC of Stock Value During Trading Day\")\n plt.show()", "_____no_output_____" ] ], [ [ "**59.** Generate a day's worth of random stock data, and aggregate / reformat it so that it has hourly summaries of the opening, highest, lowest, and closing prices", "_____no_output_____" ], [ "**60.** Now that you have your properly-formatted data, try to plot it yourself as a candlestick chart. Use the ```plot_candlestick(df)``` function above, or matplotlib's [```plot``` documentation](https://matplotlib.org/api/_as_gen/matplotlib.axes.Axes.plot.html) if you get stuck.", "_____no_output_____" ], [ "*More exercises to follow soon...*", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ] ]

[ "CC-BY-3.0" ]

[ "CC-BY-3.0" ]

[ "CC-BY-3.0" ]

[ [ [ "import matplotlib.pyplot as plt\nimport numpy as np\nimport pandas as pd\nfrom sklearn import linear_model\nimport statsmodels.api as sm\nfrom sqlalchemy import create_engine\n\n# Display preferences.\n%matplotlib inline\npd.options.display.float_format = '{:.3f}'.format\n\nimport warnings\nwarnings.filterwarnings(action=\"ignore\")\n\npostgres_user = 'dsbc_student'\npostgres_pw = '7*.8G9QH21'\npostgres_host = '142.93.121.174'\npostgres_port = '5432'\npostgres_db = 'weatherinszeged'\n\nengine = create_engine('postgresql://{}:{}@{}:{}/{}'.format(\n postgres_user, postgres_pw, postgres_host, postgres_port, postgres_db))\n\nweather_df = pd.read_sql_query('select * from weatherinszeged',con=engine)\n\n# no need for an open connection, as we're only doing a single query\nengine.dispose()", "_____no_output_____" ], [ "# Y is the target variable\nY = weather_df['apparenttemperature'] - weather_df['temperature']\n\n# X is the feature set\nX = weather_df[['humidity','windspeed']]\n\n# We add constant to the model as it's a best practice\n# to do so every time!\nX = sm.add_constant(X)\n\n# We fit an OLS model using statsmodels\nresults = sm.OLS(Y, X).fit()\n\n# We print the summary results.\nprint(results.summary())", " OLS Regression Results \n==============================================================================\nDep. Variable: y R-squared: 0.288\nModel: OLS Adj. R-squared: 0.288\nMethod: Least Squares F-statistic: 1.949e+04\nDate: Mon, 20 Jan 2020 Prob (F-statistic): 0.00\nTime: 21:09:48 Log-Likelihood: -1.7046e+05\nNo. Observations: 96453 AIC: 3.409e+05\nDf Residuals: 96450 BIC: 3.409e+05\nDf Model: 2 \nCovariance Type: nonrobust \n==============================================================================\n coef std err t P>|t| [0.025 0.975]\n------------------------------------------------------------------------------\nconst 2.4381 0.021 115.948 0.000 2.397 2.479\nhumidity -3.0292 0.024 -126.479 0.000 -3.076 -2.982\nwindspeed -0.1193 0.001 -176.164 0.000 -0.121 -0.118\n==============================================================================\nOmnibus: 3935.747 Durbin-Watson: 0.267\nProb(Omnibus): 0.000 Jarque-Bera (JB): 4613.311\nSkew: -0.478 Prob(JB): 0.00\nKurtosis: 3.484 Cond. No. 88.1\n==============================================================================\n\nWarnings:\n[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n" ], [ "#r-sqaured = .288 and Adjusted r-squared = .288\n# This is unsatisfactory because it's very low and gives a lot of room to improve the model", "_____no_output_____" ], [ "#interaction \n# Y is the target variable\nY = weather_df['apparenttemperature'] - weather_df['temperature']\n\nweather_df[\"humidity\"] = weather_df.humidity * weather_df.windspeed\n\n# X is the feature set\nX = weather_df[['humidity','windspeed']]\n\n# We add a constant to the model as it's a best practice\n# to do so every time!\nX = sm.add_constant(X)\n\n# We fit an OLS model using statsmodels\nresults = sm.OLS(Y, X).fit()\n\n# We print the summary results\nprint(results.summary())", " OLS Regression Results \n==============================================================================\nDep. Variable: y R-squared: 0.341\nModel: OLS Adj. R-squared: 0.341\nMethod: Least Squares F-statistic: 2.497e+04\nDate: Mon, 20 Jan 2020 Prob (F-statistic): 0.00\nTime: 21:11:15 Log-Likelihood: -1.6670e+05\nNo. Observations: 96453 AIC: 3.334e+05\nDf Residuals: 96450 BIC: 3.334e+05\nDf Model: 2 \nCovariance Type: nonrobust \n==============================================================================\n coef std err t P>|t| [0.025 0.975]\n------------------------------------------------------------------------------\nconst 0.2178 0.008 26.385 0.000 0.202 0.234\nhumidity -0.2854 0.002 -158.432 0.000 -0.289 -0.282\nwindspeed 0.0819 0.001 62.418 0.000 0.079 0.084\n==============================================================================\nOmnibus: 4795.516 Durbin-Watson: 0.265\nProb(Omnibus): 0.000 Jarque-Bera (JB): 9042.719\nSkew: -0.379 Prob(JB): 0.00\nKurtosis: 4.295 Cond. No. 29.5\n==============================================================================\n\nWarnings:\n[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n" ], [ "#The new r-squared and adjusted r-squared have barely risen, still low given they're below the .50 mark\n#This doesn't improve the model by much", "_____no_output_____" ], [ "# Y is the target variable\nY = weather_df['apparenttemperature'] - weather_df['temperature']\n\n# X is the feature set with visibility\nX = weather_df[['humidity','windspeed','visibility']]\n\n# We add constant to the model as it's a best practice\n# to do so every time!\nX = sm.add_constant(X)\n\n# We fit an OLS model using statsmodels\nresults = sm.OLS(Y, X).fit()\n\n# We print the summary results.\nprint(results.summary())", " OLS Regression Results \n==============================================================================\nDep. Variable: y R-squared: 0.361\nModel: OLS Adj. R-squared: 0.361\nMethod: Least Squares F-statistic: 1.815e+04\nDate: Mon, 20 Jan 2020 Prob (F-statistic): 0.00\nTime: 21:12:45 Log-Likelihood: -1.6524e+05\nNo. Observations: 96453 AIC: 3.305e+05\nDf Residuals: 96449 BIC: 3.305e+05\nDf Model: 3 \nCovariance Type: nonrobust \n==============================================================================\n coef std err t P>|t| [0.025 0.975]\n------------------------------------------------------------------------------\nconst -0.3657 0.013 -27.184 0.000 -0.392 -0.339\nhumidity -0.2603 0.002 -141.975 0.000 -0.264 -0.257\nwindspeed 0.0623 0.001 46.462 0.000 0.060 0.065\nvisibility 0.0583 0.001 54.446 0.000 0.056 0.060\n==============================================================================\nOmnibus: 4991.814 Durbin-Watson: 0.285\nProb(Omnibus): 0.000 Jarque-Bera (JB): 9859.261\nSkew: -0.378 Prob(JB): 0.00\nKurtosis: 4.371 Cond. No. 57.3\n==============================================================================\n\nWarnings:\n[1] Standard Errors assume that the covariance matrix of the errors is correctly specified.\n" ], [ "#This is the best model by far because the r-squared and adjusted r-sqaured \n#values, because it's the highest among the three at 0.361; and the AIC and BIC numbers are the lowest", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code" ] ]