-
Notifications
You must be signed in to change notification settings - Fork 0
/
kmeans.py
141 lines (107 loc) · 4.64 KB
/
kmeans.py
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
import numpy as np
import matplotlib.pyplot as plt
import math
from sklearn import datasets
import os
import shutil
from PIL import Image
import glob
centercolor = '#e60e0e'
def dist(p1, p2):
return math.sqrt((p2[0] - p1[0])**2 + (p2[1] - p1[1])**2)
def label_points(points, centers, k):
labels = np.zeros((len(points)))
for p_index in range(len(points)):
distances = [dist(points[p_index], centers[i]) for i in range(k)]
labels[p_index] = (distances.index(min(distances)))
return labels
def recompute_centers(points, labels, old_centers, k):
cluster_dict = {i:[] for i in range(k)}
for p, l in zip(points, labels):
cluster_dict[l].append(p)
centers = np.zeros((k, 2))
for i in range(k):
if len(cluster_dict[i]) > 0:
centers[i][0] = np.mean([p[0] for p in cluster_dict[i]])
centers[i][1] = np.mean([p[1] for p in cluster_dict[i]])
else:
centers[i][0] = old_centers[i][0]
centers[i][1] = old_centers[i][1]
return centers
def draw_coords_from_list(points, xrange):
move_right = xrange / 75
for p in points:
x = float(p[0])
y = float(p[1])
if int(p[0]) - x == 0:
x = int(p[0])
if int(p[1]) - y == 0:
y = int(p[1])
plt.text(p[0] + move_right, p[1], '(' + str(x) + ', ' + str(y) + ')')
def draw_coords(points, centers, xrange, plot_centers):
draw_coords_from_list(points, xrange)
if plot_centers:
draw_coords_from_list(centers, xrange)
def plot_points(points, centers, labels, xmin, xmax, ymin, ymax, filename, title, plot_centers=True, show_coords=False):
xpoints = [p[0] for p in points] + ([xmin - abs(xmin / 10)] * len(centers))
ypoints = [p[1] for p in points] + ([0] * len(centers))
xcenters = [p[0] for p in centers]
ycenters = [p[1] for p in centers]
labels = np.concatenate((labels, np.array(range(len(centers)))))
plt.xlim(xmin, xmax)
plt.ylim(ymin, ymax)
if plot_centers:
plt.scatter(xpoints, ypoints, s=40, c=labels)
plt.scatter(xcenters, ycenters, s=80, c=centercolor, marker='^')
else:
plt.scatter(xpoints, ypoints, s=40)
draw_coords(points, centers, xmax - xmin, plot_centers)
plt.title(title, loc='left')
plt.savefig(filename)
plt.clf()
def has_converged(old_centers, new_centers):
for old, new in zip(old_centers, new_centers):
if not (old == new).all():
return False
return True
def animate(points, centers, show_coords=False):
old_centers = centers + 1
pad = 1.1
xmin = np.min(points[:,0]) * pad
xmax = np.max(points[:,0]) * pad
ymin = np.min(points[:,1]) * pad
ymax = np.max(points[:,1]) * pad
k = len(centers)
if os.path.exists('frames'):
shutil.rmtree('frames')
os.makedirs('frames')
plot_points(points, centers, [], xmin, xmax, ymin, ymax, 'frames/data', 'Data Points', plot_centers=False, show_coords=show_coords)
labels = label_points(points, centers, k)
plot_points(points, centers, labels, xmin, xmax, ymin, ymax, 'frames/0', 'Epoch: 1')
index = 1
epoch = 1
labels = label_points(points, centers, k)
while not has_converged(old_centers, centers):
old_centers = centers
centers = recompute_centers(points, labels, old_centers, k)
plot_points(points, centers, labels, xmin, xmax, ymin, ymax, 'frames/' + str(index), 'Epoch: ' + str(epoch), show_coords=show_coords)
index += 1
labels = label_points(points, centers, k)
plot_points(points, centers, labels, xmin, xmax, ymin, ymax, 'frames/' + str(index), 'Epoch: ' + str(epoch), show_coords=show_coords)
index += 1
plot_points(points, centers, labels, xmin, xmax, ymin, ymax, 'frames/' + str(index), 'Epoch: ' + str(epoch), show_coords=show_coords)
index += 1
epoch += 1
for i in range(6):
plot_points(points, centers, labels, xmin, xmax, ymin, ymax, 'frames/' + str(index + i), 'Epoch: ' + str(epoch - 1) + ' '*8 + 'Done!')
frames = []
images = sorted(glob.glob('frames/*.png'), key=os.path.getmtime)
for img in images:
frames.append(Image.open(img))
frames[0].save('kmeans_animation.gif', format='GIF', append_images=frames[1:], save_all=True, duration=500, loop=0)
# shutil.rmtree('frames')
# points = np.array([(4, 4), (-3, 2), (-1, -1), (2, 6), (-3, -3), (1, -5)])
# centers = np.array([[0, -1], [0, 1]])
points, labels = datasets.make_blobs(n_samples=500, n_features=2, centers=5, random_state=2)
centers = np.array([[0, -1], [0, 1], [4, -1], [2, 3], [5, -4], [3, 3]])
animate(points, centers, show_coords=False)