File size: 12,748 Bytes
674a23c c37bb09 674a23c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 |
from typing import Callable, List, Tuple
import numpy as np
import pandas as pd
from gensim.models.doc2vec import Doc2Vec, TaggedDocument
import tensorflow as tf
from tensorflow import keras
from keras.preprocessing.text import Tokenizer
def read_data(filepath="./csvs/"):
"""
Reading CSV files of the dataset.
Parameters:
----------
- filepath : str
Defines the path that contains the CSV files.
Returns:
--------
A tuple contains the following:
- X_train : pd.DataFrame
- y_train : pd.Series
- X_test : pd.DataFrame
- y_test : pd.Series
"""
X_train = pd.read_csv(filepath + "X_train.csv")
X_train = X_train.iloc[:, 1:]
X_test = pd.read_csv(filepath + "X_test.csv")
X_test = X_test.iloc[:, 1:]
y_train = pd.read_csv(filepath + "y_train.csv")
y_train = y_train.iloc[:, 1:]
y_test = pd.read_csv(filepath + "y_test.csv")
y_test = y_test.iloc[:, 1:]
return X_train, X_test, y_train, y_test
def train_model(
model_building_func: Callable[[], keras.models.Sequential],
X_train_vectors: pd.DataFrame | np.ndarray | tf.Tensor,
y_train: pd.Series,
k: int = 4,
num_epochs: int = 30,
batch_size: int = 64,
) -> Tuple[
List[keras.models.Sequential],
List[List[float]],
List[List[float]],
List[List[float]],
List[List[float]],
]:
"""
Trains a model on `X_train_vectors` and `y_train` using k-fold cross-validation.
Parameters:
-----------
- model_building_func : Callable[[], tf.keras.models.Sequential]
A function that builds and compiles a Keras Sequential model.
- X_train_vectors : pd.DataFrame
The training input data.
- y_train : pd.Series
The training target data.
- k : int, optional
The number of folds for cross-validation (default is 4).
- num_epochs : int, optional
The number of epochs to train for (default is 30).
- batch_size : int, optional
The batch size to use during training (default is 64).
Returns:
--------
A tuple containing the following items:
- all_models : List[keras.models.Sequential]
A list of `k` trained models.
- all_losses : List[List[float]]
A `k` by `num_epochs` list containing the training losses for each fold.
- all_val_losses : List[List[float]]
A `k` by `num_epochs` list containing the validation losses for each fold.
- all_acc : List[List[float]]
A `k` by `num_epochs` list containing the training accuracies for each fold.
- all_val_acc : List[List[float]]
A `k` by `num_epochs` list containing the validation accuracies for each fold.
"""
num_validation_samples = len(X_train_vectors) // k
all_models = []
all_losses = []
all_val_losses = []
all_accuracies = []
all_val_accuracies = []
for fold in range(k):
print(f"fold: {fold+1}")
validation_data = X_train_vectors[
num_validation_samples * fold : num_validation_samples * (fold + 1)
]
validation_targets = y_train[
num_validation_samples * fold : num_validation_samples * (fold + 1)
]
training_data = np.concatenate(
[
X_train_vectors[: num_validation_samples * fold],
X_train_vectors[num_validation_samples * (fold + 1) :],
]
)
training_targets = np.concatenate(
[
y_train[: num_validation_samples * fold],
y_train[num_validation_samples * (fold + 1) :],
]
)
model = model_building_func()
history = model.fit(
training_data,
training_targets,
validation_data=(validation_data, validation_targets),
epochs=num_epochs,
batch_size=batch_size,
)
all_models.append(model)
all_losses.append(history.history["loss"])
all_val_losses.append(history.history["val_loss"])
all_accuracies.append(history.history["accuracy"])
all_val_accuracies.append(history.history["val_accuracy"])
return (all_models, all_losses, all_val_losses, all_accuracies, all_val_accuracies)
def print_testing_loss_accuracy(
all_models: List[keras.models.Sequential],
X_test_vectors: pd.DataFrame | np.ndarray | tf.Tensor,
y_test: pd.Series,
) -> None:
"""
Displaying testing loss and testing accuracy of each model in `all_models`,
and displaying their average.
Parameters:
------------
- all_models : List[keras.models.Sequential]
A list of size `k` contains trained models.
- X_test_vectors : pd.DataFrame
Contains testing vectors.
- y_test : pd.Series
Contains testing labels.
"""
sum_testing_losses = 0.0
sum_testing_accuracies = 0.0
for i, model in enumerate(all_models):
print(f"model: {i+1}")
loss_accuracy = model.evaluate(X_test_vectors, y_test, verbose=1)
sum_testing_losses += loss_accuracy[0]
sum_testing_accuracies += loss_accuracy[1]
print("====" * 20)
num_models = len(all_models)
avg_testing_loss = sum_testing_losses / num_models
avg_testing_acc = sum_testing_accuracies / num_models
print(f"average testing loss: {avg_testing_loss:.3f}")
print(f"average testing accuracy: {avg_testing_acc:.3f}")
def calculate_average_measures(
all_losses: list[list[float]],
all_val_losses: list[list[float]],
all_accuracies: list[list[float]],
all_val_accuracies: list[list[float]],
) -> Tuple[
List[keras.models.Sequential],
List[List[float]],
List[List[float]],
List[List[float]],
List[List[float]],
]:
"""
Calculate the average measures of cross-validated results.
Parameters:
------------
- all_losses : List[List[float]]
A `k` by `num_epochs` list contains the values of training losses.
- all_val_losses : List[List[float]]
A `k` by `num_epochs` list contains the values of validation losses.
- all_accuracies : List[List[float]]
A `k` by `num_epochs` list contains the values of training accuracies.
- all_val_accuracies : List[List[float]]
A `k` by `num_epochs` list contains the values of validation accuracies.
Returns:
--------
A tuple containing the following items:
- avg_loss_hist : List[float]
A list of length `num_epochs` contains the average of training losses.
- avg_val_loss_hist : List[float]
A list of length `num_epochs` contains the average of validaton losses.
- avg_acc_hist : List[float]
A list of length `num_epochs` contains the average of training accuracies.
- avg_val_acc_hist : List[float]
A list of length `num_epochs` contains the average of validation accuracies.
"""
num_epochs = len(all_losses[0])
avg_loss_hist = [np.mean([x[i] for x in all_losses]) for i in range(num_epochs)]
avg_val_loss_hist = [
np.mean([x[i] for x in all_val_losses]) for i in range(num_epochs)
]
avg_acc_hist = [np.mean([x[i] for x in all_accuracies]) for i in range(num_epochs)]
avg_val_acc_hist = [
np.mean([x[i] for x in all_val_accuracies]) for i in range(num_epochs)
]
return (avg_loss_hist, avg_val_loss_hist, avg_acc_hist, avg_val_acc_hist)
class Doc2VecModel:
"""Responsible of creating, initializing, and training Doc2Vec embeddings model."""
def __init__(self, vector_size=50, min_count=2, epochs=100, dm=1, window=5) -> None:
"""
Initalize a Doc2Vec model.
Parameters:
------------
- vector_size : int, optional
Dimensionality of the feature vectors (Default is 50).
- min_count : int, optional
Ignores all words with total frequency lower than this (Default is 2).
- epochs : int, optional
Represents the number of training epochs (Default is 100).
- dm : int, optional
Defines the training algorithm. If `dm=1`, 'distributed memory' (PV-DM) is used.
Otherwise, `distributed bag of words` (PV-DBOW) is employed (Default is 1).
- window : int, optional
The maximum distance between the current and predicted word within a
sentence (Default is 5).
"""
self.doc2vec_model = Doc2Vec(
vector_size=vector_size,
min_count=min_count,
epochs=epochs,
dm=dm,
seed=865,
window=window,
)
def train_doc2vec_embeddings_model(
self, tagged_docs_train: List[TaggedDocument]
) -> Doc2Vec:
"""
Train Doc2Vec model on `tagged_docs_train`.
Parameters:
------------
- tagged_docs_train : list[TaggedDocument]
Contains the required format of training Doc2Vec model.
Returns:
--------
- doc2vec_model : Doc2Vec
The trained Doc2Vec model.
"""
self.doc2vec_model.build_vocab(tagged_docs_train)
self.doc2vec_model.train(
tagged_docs_train,
total_examples=self.doc2vec_model.corpus_count,
epochs=self.doc2vec_model.epochs,
)
return self.doc2vec_model
class GloveModel:
"""Responsible for creating and generating the glove embedding layer"""
def __init__(self) -> None:
pass
def _generate_glove_embedding_index(
self, glove_file_path: str = "GloVe/glove.6B.50d.txt"
) -> dict:
"""
Responsible for generating glove embedding index.
Parameters:
------------
- glove_file_path : str
Defines the path of the pretrained GloVe embeddings text file
(Default is "GloVe/glove.6B.50d.txt").
Returns:
--------
- embedding_index : dict
Contains each word as a key, and its co-effeicents as a value.
"""
embeddings_index = {}
with open(glove_file_path, encoding="utf8") as f:
for line in f:
values = line.split()
word = values[0]
coefs = np.asarray(values[1:], dtype="float32")
embeddings_index[word] = coefs
return embeddings_index
def _generate_glove_embedding_matrix(
self, word_index: dict, embedding_index: dict, max_length: int
) -> np.ndarray:
"""
Generating embedding matrix of each word in `word_index`.
Parameters:
-----------
- word_index : dict
Contains words as keys with there indicies as values.
- embedding_index : dict
Contains each word as a key, and its co-effeicents as a value.
- max_length : int
Defines the size of the embedding vector of each word in the
embedding matrix.
Returns:
--------
- embedding_matrix : np.ndarray
Contains all embedding vectors for each word in`word_index`.
"""
embedding_matrix = np.zeros((len(word_index) + 1, max_length))
for word, i in word_index.items():
embedding_vector = embedding_index.get(word)
if embedding_vector is not None:
embedding_matrix[i] = embedding_vector
return embedding_matrix
def generate_glove_embedding_layer(
self, glove_tokenizer: Tokenizer, max_length: int = 50
) -> keras.layers.Embedding:
"""
Create GloVe embedding layer for later usage in the neural network.
Paramters:
----------
- glove_tokenizer : Tokenizer
Trained tokenizer on training data to extract word index from it.
- max_length : int, optional
Defines the maximum length of the output embedding vector for
each word. (Default is 50).
Returns:
--------
- embedding_layer : keras.layers.Embedding
An embedding layer of size `word index + 1` by `max_length` with
trained weights that can be used a vectorizer of case facts.
"""
word_index = glove_tokenizer.word_index
embedding_index = self._generate_glove_embedding_index()
embedding_matrix = self._generate_glove_embedding_matrix(
word_index, embedding_index, max_length
)
embedding_layer = keras.layers.Embedding(
len(word_index) + 1,
max_length,
weights=[embedding_matrix],
input_length=max_length,
trainable=False,
)
return embedding_layer
|