mervenoyan
commited on
Commit
•
75df68a
1
Parent(s):
4bfd1a6
initial commit
Browse files- app.py +169 -0
- requirements.txt +2 -0
app.py
ADDED
@@ -0,0 +1,169 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
import time
|
3 |
+
import warnings
|
4 |
+
|
5 |
+
from functools import partial
|
6 |
+
import gradio as gr
|
7 |
+
import numpy as np
|
8 |
+
import matplotlib.pyplot as plt
|
9 |
+
|
10 |
+
from sklearn import cluster, datasets
|
11 |
+
from sklearn.preprocessing import StandardScaler
|
12 |
+
from itertools import cycle, islice
|
13 |
+
|
14 |
+
|
15 |
+
|
16 |
+
def train_models(selected_data, n_samples, n_clusters, n_neighbors, cls_name):
|
17 |
+
np.random.seed(0)
|
18 |
+
default_base = {"n_neighbors": 10, "n_clusters": 3}
|
19 |
+
noisy_circles = datasets.make_circles(n_samples=n_samples, factor=0.5, noise=0.05)
|
20 |
+
noisy_moons = datasets.make_moons(n_samples=n_samples, noise=0.05)
|
21 |
+
blobs = datasets.make_blobs(n_samples=n_samples, random_state=8)
|
22 |
+
no_structure = np.random.rand(n_samples, 2), None
|
23 |
+
|
24 |
+
# Anisotropicly distributed data
|
25 |
+
random_state = 170
|
26 |
+
X, y = datasets.make_blobs(n_samples=n_samples, random_state=random_state)
|
27 |
+
transformation = [[0.6, -0.6], [-0.4, 0.8]]
|
28 |
+
X_aniso = np.dot(X, transformation)
|
29 |
+
aniso = (X_aniso, y)
|
30 |
+
|
31 |
+
# blobs with varied variances
|
32 |
+
varied = datasets.make_blobs(
|
33 |
+
n_samples=n_samples, cluster_std=[1.0, 2.5, 0.5], random_state=random_state
|
34 |
+
)
|
35 |
+
|
36 |
+
dataset_list = {
|
37 |
+
"Noisy Circles": [noisy_circles, {"n_clusters": n_clusters}],
|
38 |
+
"Noisy Moons": [noisy_moons, {"n_clusters": n_clusters}],
|
39 |
+
"Varied": [varied, {"n_neighbors": n_neighbors}],
|
40 |
+
"Aniso": [aniso, {"n_neighbors": n_neighbors}],
|
41 |
+
"Blobs": [blobs, {}],
|
42 |
+
"No Structure": [no_structure, {}],
|
43 |
+
}
|
44 |
+
|
45 |
+
params = default_base.copy()
|
46 |
+
params.update(dataset_list[selected_data][1])
|
47 |
+
|
48 |
+
X, y = dataset_list[selected_data][0]
|
49 |
+
|
50 |
+
X = StandardScaler().fit_transform(X)
|
51 |
+
|
52 |
+
ward = cluster.AgglomerativeClustering(
|
53 |
+
n_clusters=params["n_clusters"], linkage="ward"
|
54 |
+
)
|
55 |
+
complete = cluster.AgglomerativeClustering(
|
56 |
+
n_clusters=params["n_clusters"], linkage="complete"
|
57 |
+
)
|
58 |
+
average = cluster.AgglomerativeClustering(
|
59 |
+
n_clusters=params["n_clusters"], linkage="average"
|
60 |
+
)
|
61 |
+
single = cluster.AgglomerativeClustering(
|
62 |
+
n_clusters=params["n_clusters"], linkage="single"
|
63 |
+
)
|
64 |
+
|
65 |
+
clustering_algorithms = {
|
66 |
+
"Single Linkage": single,
|
67 |
+
"Average Linkage": average,
|
68 |
+
"Complete Linkage": complete,
|
69 |
+
"Ward Linkage": ward,
|
70 |
+
}
|
71 |
+
|
72 |
+
t0 = time.time()
|
73 |
+
algorithm = clustering_algorithms[cls_name]
|
74 |
+
# catch warnings related to kneighbors_graph
|
75 |
+
with warnings.catch_warnings():
|
76 |
+
warnings.filterwarnings(
|
77 |
+
"ignore",
|
78 |
+
message="the number of connected components of the "
|
79 |
+
+ "connectivity matrix is [0-9]{1,2}"
|
80 |
+
+ " > 1. Completing it to avoid stopping the tree early.",
|
81 |
+
category=UserWarning,
|
82 |
+
)
|
83 |
+
algorithm.fit(X)
|
84 |
+
|
85 |
+
t1 = time.time()
|
86 |
+
if hasattr(algorithm, "labels_"):
|
87 |
+
y_pred = algorithm.labels_.astype(int)
|
88 |
+
else:
|
89 |
+
y_pred = algorithm.predict(X)
|
90 |
+
|
91 |
+
fig, ax = plt.subplots()
|
92 |
+
|
93 |
+
colors = np.array(
|
94 |
+
list(
|
95 |
+
islice(
|
96 |
+
cycle(
|
97 |
+
[
|
98 |
+
"#377eb8",
|
99 |
+
"#ff7f00",
|
100 |
+
"#4daf4a",
|
101 |
+
"#f781bf",
|
102 |
+
"#a65628",
|
103 |
+
"#984ea3",
|
104 |
+
"#999999",
|
105 |
+
"#e41a1c",
|
106 |
+
"#dede00",
|
107 |
+
]
|
108 |
+
),
|
109 |
+
int(max(y_pred) + 1),
|
110 |
+
)
|
111 |
+
)
|
112 |
+
)
|
113 |
+
ax.scatter(X[:, 0], X[:, 1], color=colors[y_pred])
|
114 |
+
|
115 |
+
ax.set_xlim(-2.5, 2.5)
|
116 |
+
ax.set_ylim(-2.5, 2.5)
|
117 |
+
ax.set_xticks(())
|
118 |
+
ax.set_yticks(())
|
119 |
+
|
120 |
+
return fig
|
121 |
+
|
122 |
+
|
123 |
+
def iter_grid(n_rows, n_cols):
|
124 |
+
# create a grid using gradio Block
|
125 |
+
for _ in range(n_rows):
|
126 |
+
with gr.Row():
|
127 |
+
for _ in range(n_cols):
|
128 |
+
with gr.Column():
|
129 |
+
yield
|
130 |
+
|
131 |
+
title = "Compare linkages in hierarchical clustering"
|
132 |
+
with gr.Blocks(title=title) as demo:
|
133 |
+
gr.Markdown(f"## {title}")
|
134 |
+
gr.Markdown("This app demonstrates different linkage methods in"
|
135 |
+
" hierarchical clustering")
|
136 |
+
|
137 |
+
|
138 |
+
input_models = ["Single Linkage", "Average Linkage", "Complete Linkage",
|
139 |
+
"Ward Linkage"]
|
140 |
+
input_data = gr.Radio(
|
141 |
+
choices=["Noisy Circles", "Noisy Moons",
|
142 |
+
"Varied", "Aniso", "Blobs", "No Structure"],
|
143 |
+
value="Noisy Moons"
|
144 |
+
)
|
145 |
+
n_samples = gr.Slider(minimum=500, maximum=2000, step=50,
|
146 |
+
label = "Number of Samples")
|
147 |
+
|
148 |
+
n_neighbors = gr.Slider(minimum=2, maximum=5, step=1,
|
149 |
+
label = "Number of neighbors")
|
150 |
+
n_clusters = gr.Slider(minimum=2, maximum=5, step=1,
|
151 |
+
label = "Number of Clusters")
|
152 |
+
counter = 0
|
153 |
+
|
154 |
+
for _ in iter_grid(2, 5):
|
155 |
+
if counter >= len(input_models):
|
156 |
+
break
|
157 |
+
|
158 |
+
input_model = input_models[counter]
|
159 |
+
plot = gr.Plot(label=input_model)
|
160 |
+
fn = partial(train_models, cls_name=input_model)
|
161 |
+
input_data.change(fn=fn, inputs=[input_data, n_samples, n_clusters, n_neighbors], outputs=plot)
|
162 |
+
n_samples.change(fn=fn, inputs=[input_data, n_samples, n_clusters, n_neighbors], outputs=plot)
|
163 |
+
|
164 |
+
n_neighbors.change(fn=fn, inputs=[input_data, n_samples, n_clusters, n_neighbors], outputs=plot)
|
165 |
+
n_clusters.change(fn=fn, inputs=[input_data, n_samples, n_clusters, n_neighbors], outputs=plot)
|
166 |
+
counter += 1
|
167 |
+
|
168 |
+
|
169 |
+
demo.launch()
|
requirements.txt
ADDED
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
1 |
+
scikit-learn
|
2 |
+
matplotlib
|