-
Notifications
You must be signed in to change notification settings - Fork 0
/
MiniBatch.py
62 lines (46 loc) · 1.52 KB
/
MiniBatch.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
import numpy as np
import matplotlib.pyplot as plt
class MiniBatchKMeans():
def __init__(self, x, k, b):
self.x = x
self.k = k
self.b = b
self.N, self.ndim = x.shape
self.initialise_cSv()
self.convCritirion = False
def initialise_cSv(self):
# Algorithm 2
choice = np.random.choice(self.N, self.k)
self.c = self.x[choice]
self.S = self.x[choice]
self.v = np.ones([self.k, 1], dtype=np.int32)
return self.c, self.S, self.v
def a(self, i):
# argmin of euclidean distance
return (np.linalg.norm(self.c.reshape(-1, self.ndim) - self.x[i], axis=1)).argmin()
def accumulate(self, i):
# Algorithm 3
am = self.a(i)
self.S[am] += self.x[i]
self.v[am] += 1
def mbatch(self, max_iter):
# Algorithm 4
iter = 0
while not self.convCritirion and iter < max_iter:
temp = self.c
M = np.random.choice(self.N, self.b)
for i in M:
self.accumulate(self.a(i))
self.c = self.S / self.v.reshape(-1, 1)
if (temp - self.c == 0).all():
self.convCritirion = True
iter += 1
def show(self, keep = False):
C = [self.a(i) for i in range(self.N)]
plt.cla()
plt.scatter(self.x[:,0], self.x[:,1], c=C)
plt.plot(self.c[:,0],self.c[:,1],'x', color="black",markersize=10)
plt.draw()
if keep :
plt.ioff()
plt.show()