/
k_means.py
executable file
·348 lines (306 loc) · 15.1 KB
/
k_means.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
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
import numpy as np
import timeit
import time
from dunn import *
from sklearn.cluster import KMeans
debug = False
def crossovers_centres(solution_set, k, dim, points, m, n = 2):
""" 2*m indicates the number of solutions that need to be generated from the crossover.
n is the number of crossovers"""
if (2*m)>solution_set.shape[0]:
print("Error!")
return
combinations = np.linspace(0,solution_set.shape[0]-1,solution_set.shape[0], dtype = int)
np.random.shuffle(combinations)
combinations = combinations[:2*m]
debug and print("Combinations", combinations)
new_solutions = np.zeros(shape = (2*m, k, dim))
for i in range(m):
gene1 = solution_set[combinations[2*i]].reshape(k*dim)
gene2 = solution_set[combinations[2*i+1]].reshape(k*dim)
crossover_points = np.sort(np.random.randint(1, k*dim, size = n))
gene_1_new = np.zeros(shape = (1, k*dim))
gene_2_new = np.zeros(shape = (1, k*dim))
debug and print("Crossover", crossover_points)
gene_1_new[0][:crossover_points[0]] = gene2[:crossover_points[0]]
gene_2_new[0][:crossover_points[0]] = gene1[:crossover_points[0]]
for j in range(1,n):
if j%2==0:
gene_2_new[0][crossover_points[j-1]:crossover_points[j]] = gene1[crossover_points[j-1]:crossover_points[j]]
gene_1_new[0][crossover_points[j-1]:crossover_points[j]] = gene2[crossover_points[j-1]:crossover_points[j]]
else:
gene_1_new[0][crossover_points[j-1]:crossover_points[j]] = gene1[crossover_points[j-1]:crossover_points[j]]
gene_2_new[0][crossover_points[j-1]:crossover_points[j]] = gene2[crossover_points[j-1]:crossover_points[j]]
if n==1:
gene_2_new[0][crossover_points[-1]:] = gene2[crossover_points[-1]:]
gene_1_new[0][crossover_points[-1]:] = gene1[crossover_points[-1]:]
else:
if n%2==1:
gene_2_new[0][crossover_points[-1]:] = gene1[crossover_points[-1]:]
gene_1_new[0][crossover_points[-1]:] = gene2[crossover_points[-1]:]
else:
gene_2_new[0][crossover_points[-1]:] = gene1[crossover_points[-1]:]
gene_1_new[0][crossover_points[-1]:] = gene2[crossover_points[-1]:]
debug and print("gene1", gene1)
debug and print("gene1", gene2)
debug and print("gene1_u", gene_1_new)
debug and print("gene2_u", gene_2_new)
new_solutions[2*i] = np.copy(gene_1_new[0].reshape(k, dim))
new_solutions[2*i+1] = np.copy(gene_2_new[0].reshape(k, dim))
return np.copy(new_solutions)
def mutations_centres(solution_set, k, dim, points, p = 0.1):
for i in range(solution_set.shape[0]):
if np.random.rand()<=p:
debug and print("Yes")
position = np.random.randint(0, k)
position2 = np.random.randint(0, dim)
debug and print("Position is ", position)
solution_set[i][position][position2] = np.random.rand()
return np.copy(solution_set)
def k_means_genetic_generation_centres(solution_set, datapoints, k, dim, points, top_n, crossover_points, p, allowed_error):
new_solution = np.zeros(shape = solution_set.shape)
scores = np.zeros(shape = (solution_set.shape[0]))
for i in range(solution_set.shape[0]):
centres = np.copy(solution_set[i])
clusters = reassignPoints(datapoints, np.copy(centres), points, k, dim)
scores[i] = calculateError(datapoints, np.copy(clusters), np.copy(centres), k , dim)
top_n_indices = scores.argsort()[:top_n]
if scores[top_n_indices[0]]<allowed_error:
return (True, solution_set, top_n_indices[0], scores[top_n_indices[0]])
for i in range(top_n):
new_solution[i] = np.copy(solution_set[top_n_indices[i]])
crossover_considerations = np.zeros(shape = (solution_set.shape[0] - top_n, k, dim))
top_n_indices = scores.argsort()[top_n:]
for i in range(solution_set.shape[0] - top_n):
crossover_considerations[i] = np.copy(solution_set[top_n_indices[i]])
final_crossover = crossovers_centres(np.copy(crossover_considerations), k, dim, points, int((solution_set.shape[0] - top_n)/2), n = crossover_points)
mutated = mutations_centres(np.copy(final_crossover), k, dim, points, p)
top_n_indices = scores.argsort()[:top_n]
for i in range(solution_set.shape[0] - top_n):
new_solution[top_n + i] = np.copy(mutated[i])
return (False, np.copy(new_solution), top_n_indices[0], scores[top_n_indices[0]])
def k_means_genetic_centres(datapoints, dim, points, k, solution_set_size = 100, top_n = 30, crossover_points = 2, p = 0.1, allowed_error = 10, max_generations = 10):
generation = 0
solution_set = np.random.rand(solution_set_size, k, dim)
while True:
x = k_means_genetic_generation_centres(np.copy(solution_set), datapoints, k, dim, points, top_n, crossover_points, p, allowed_error)
if x[0]:
break
else:
solution_set = np.copy(x[1])
debug and print(x[3], x[2])
debug and print(solution_set)
generation+=1
if generation>max_generations:
break
debug and print(x[1][x[2]])
debug and print("Error = ", x[2])
return x
def k_means_genetic_generation(solution_set, datapoints, k, dim, points, top_n, new_n, crossover_points, p, allowed_error):
new_solution = np.zeros(shape = solution_set.shape, dtype = np.int32)
scores = np.zeros(shape = (solution_set.shape[0]))
for i in range(solution_set.shape[0]):
solution = np.copy(solution_set[i])
centres = findCentres(datapoints, np.copy(solution), k, dim)
scores[i] = calculateError(datapoints, np.copy(solution), np.copy(centres), k , dim)
# print(scores, np.argmin(scores),scores.argsort()[0])
top_n_indices = scores.argsort()[:top_n]
if scores[top_n_indices[0]]<allowed_error:
return (True, solution_set, top_n_indices[0], scores[top_n_indices[0]])
for i in range(top_n):
new_solution[i] = np.copy(solution_set[top_n_indices[i]])
new_solution[top_n:top_n+new_n, :] = np.random.randint(0, k, size = (new_n, points))
crossover_considerations = np.zeros(shape = (solution_set.shape[0] - top_n - new_n, points))
top_n_indices = scores.argsort()[top_n:]
for i in range(solution_set.shape[0] - top_n - new_n):
crossover_considerations[i] = np.copy(solution_set[top_n_indices[i]])
final_crossover = crossovers(np.copy(crossover_considerations), k, dim, points, int((solution_set.shape[0] - top_n - new_n)/2), n = crossover_points)
mutated = mutations(np.copy(final_crossover), k, dim, points, p)
top_n_indices = scores.argsort()[:top_n]
for i in range(solution_set.shape[0] - top_n - new_n):
new_solution[top_n + new_n + i] = np.copy(mutated[i])
return (False, np.copy(new_solution), top_n_indices[0], scores[top_n_indices[0]])
def compare(array1, array2):
if array1.shape != array2.shape:
return False
else:
for i in range(array1.shape[0]):
#for j in range(array1.shape[1]):
if array1[i]!=array2[i]:
return False
return True
def k_means_genetic(datapoints, dim, points, k, solution_set_size = 100, top_n = 30, new_n = 40, crossover_points = 1, p = 0.1, allowed_error = 10, max_generations = 10):
solution_set = np.random.randint(0, k, size = (solution_set_size, points))
debug and print(solution_set)
generation = 0
debug and print("Generation", generation, "starting")
x = None
while True:
x = k_means_genetic_generation(np.copy(solution_set), datapoints, k, dim, points, top_n, new_n, crossover_points, p, allowed_error)
if x[0]:
break
else:
debug and print(x[1])
solution_set = np.copy(x[1])
debug and print(solution_set)
generation+=1
if generation>max_generations:
break
return x
def crossovers(solution_set, k, dim, points, m, n = 2):
""" 2*m indicates the number of solutions that need to be generated from the crossover.
n is the number of crossovers"""
if (2*m)>solution_set.shape[0]:
print("Error!")
return
combinations = np.linspace(0,solution_set.shape[0]-1,solution_set.shape[0], dtype = int)
np.random.shuffle(combinations)
combinations = combinations[:2*m]
debug and print("Combinations", combinations)
new_solutions = np.zeros(shape = (2*m, points))
for i in range(m):
gene1 = solution_set[combinations[2*i]]
gene2 = solution_set[combinations[2*i+1]]
crossover_points = np.sort(np.random.randint(1, points-1, size = n))
gene_1_new = np.zeros(shape = (1,points))
gene_2_new = np.zeros(shape = (1,points))
debug and print("Crossover", crossover_points)
gene_1_new[0][:crossover_points[0]] = gene2[:crossover_points[0]]
gene_2_new[0][:crossover_points[0]] = gene1[:crossover_points[0]]
for j in range(1,n):
if j%2==0:
gene_2_new[0][crossover_points[j-1]:crossover_points[j]] = gene1[crossover_points[j-1]:crossover_points[j]]
gene_1_new[0][crossover_points[j-1]:crossover_points[j]] = gene2[crossover_points[j-1]:crossover_points[j]]
else:
gene_1_new[0][crossover_points[j-1]:crossover_points[j]] = gene1[crossover_points[j-1]:crossover_points[j]]
gene_2_new[0][crossover_points[j-1]:crossover_points[j]] = gene2[crossover_points[j-1]:crossover_points[j]]
if n==1:
gene_2_new[0][crossover_points[-1]:] = gene2[crossover_points[-1]:]
gene_1_new[0][crossover_points[-1]:] = gene1[crossover_points[-1]:]
else:
if n%2==1:
gene_2_new[0][crossover_points[-1]:] = gene1[crossover_points[-1]:]
gene_1_new[0][crossover_points[-1]:] = gene2[crossover_points[-1]:]
else:
gene_2_new[0][crossover_points[-1]:] = gene1[crossover_points[-1]:]
gene_1_new[0][crossover_points[-1]:] = gene2[crossover_points[-1]:]
debug and print("gene1", gene1)
debug and print("gene1", gene2)
debug and print("gene1_u", gene_1_new)
debug and print("gene2_u", gene_2_new)
new_solutions[2*i][:] = np.copy(gene_1_new.astype(np.int32, copy = False))
new_solutions[2*i+1][:] = np.copy(gene_2_new.astype(np.int32, copy = False))
return np.copy(new_solutions)
def mutations(solution_set, k, dim, points, p = 0.1, number_in_each = None):
master_array = np.arange(points)
if number_in_each is None:
number_in_each = int(points*0.01)
for i in range(solution_set.shape[0]):
if np.random.rand()<=p:
debug and print("Yes")
np.random.shuffle(master_array)
position = np.random.choice(master_array, number_in_each)
debug and print("Position is ", position)
for each in position:
solution_set[i][each] = np.random.randint(0,k)
return np.copy(solution_set)
def findCentres(datapoints, solution, k, dim):
centres = np.random.rand(k, dim)
total = 0
for i in range(k):
temp = (datapoints[solution==i,:])
total += temp.size/dim
if temp.size>0:
centres[i,:] = np.mean(temp, axis = 0)
else:
centres[i,:] = np.random.randint(1, size = dim)
return np.copy(centres)
def calculateError(datapoints, solution, centres, k , dim):
error = 0.0
for i in range(k):
temp = (datapoints[solution == i, :] - centres[i,:])
temp = np.multiply(temp, temp)
error += np.sum(np.sqrt(np.sum(temp, axis = 1)), axis = 0)
return error
def reassignPoints(datapoints, centres, points, k, dim):
distance = np.zeros(shape=(points, k))
for i in range(k):
distance[:,i] = np.sqrt(np.sum(np.multiply(datapoints - centres[i,:],datapoints - centres[i,:]), axis=1))
temp3 = np.argmin(distance, axis = 1)
return np.copy(temp3)
def k_means_regular(datapoints, points, dim, k, allowed_error = 10, max_generations = 10, initial_solution = None):
solution = np.random.randint(k, size = points)
if initial_solution is not None:
solution = np.copy(initial_solution)
centres = findCentres(datapoints, solution, k, dim)
error = calculateError(datapoints, solution, centres, k, dim)
generation = 0
while error>allowed_error and generation<max_generations:
solution = reassignPoints(datapoints.copy(), centres.copy(), points, k, dim)
centres = findCentres(datapoints.copy(), solution.copy(), k, dim)
error = calculateError(datapoints.copy(), solution.copy(), centres.copy(), k, dim)
generation+=1
debug and print(generation, error)
return (solution, error, generation)
def paper_approach_2(datapoints, dim, points, k, solution_set_size = 100, top_n = 30, new_n = 40, crossover_points = 1, p = 0.1, allowed_error = 10, initial_generations = 10, max_iteration = 300):
"""http://ir.kaist.ac.kr/anthology/2009.06-Al-Shboul.pdf"""
""" Genetic Algorithm used to initialize the solution of K-means algorithm."""
s = k_means_genetic(datapoints, dim, points, k, solution_set_size = solution_set_size, top_n = top_n, new_n = new_n, crossover_points = crossover_points, p = p, allowed_error = allowed_error, max_generations = initial_generations)
best_solution = s[1][s[2]]
return k_means_regular(datapoints, points, dim, k, allowed_error = 0.001, max_generations = max_iteration, initial_solution = best_solution)
def paper_approach(datapoints, dim, points, k, solution_set_size = 100, top_n = 30, new_n = 40, crossover_points = 1, p = 0.1, allowed_error = 10, max_generations = 10, initial_iteration = 2):
"""http://ir.kaist.ac.kr/anthology/2009.06-Al-Shboul.pdf"""
""" K-means algorithm is used to initialize the solution set of the Genetic Algorithm."""
solution_set = np.zeros(shape = (solution_set_size, points))
for i in range(solution_set_size):
solution_set[i] = k_means_regular(datapoints, points, dim, k, allowed_error = 0.001, max_generations = initial_iteration)[0]
generation = 0
x = None
while True:
x = k_means_genetic_generation(np.copy(solution_set), datapoints, k, dim, points, top_n, new_n, crossover_points, p, allowed_error)
if x[0]:
break
else:
debug and print(x[1])
solution_set = np.copy(x[1])
debug and print(x[3], x[2])
debug and print(solution_set)
generation+=1
if generation>max_generations:
break
# print(x[1][x[2]])
# print("Error = ", x[3])
return x
#Initial data which is provided
k = 10
points = 10
dim = 25
datapoints = np.random.rand(points, dim)
#s = k_means_regular(datapoints, points, dim, k, allowed_error = 0.001, max_generations = 500)
f = open("readings2.txt", "w")
for i in range(2):
datapoints = np.random.rand(points, dim)
start = time.time()
r = k_means_regular(datapoints, points, dim, k, allowed_error = 0.001, max_generations = 500)
regt = time.time() - start
start = time.time()
g = k_means_genetic(datapoints, dim, points, k, solution_set_size = 100, top_n = 20, new_n = 10, crossover_points = 2, p = 0.15, allowed_error = 0.001, max_generations = 50)
gent = time.time() - start
start = time.time()
pa1 = paper_approach(datapoints, dim, points, k, solution_set_size = 100, top_n = 20, new_n = 0, crossover_points = 2, p = 0.15, allowed_error = 21.5, max_generations = 20, initial_iteration = 5)
pa1t = time.time() - start
start = time.time()
pa2 = paper_approach_2(datapoints, dim, points, k, solution_set_size = 100, top_n = 20, new_n = 0, crossover_points = 1, p = 0.15, allowed_error = 10, initial_generations = 3, max_iteration = 100)
pa2t = time.time() - start
lll = [i, dim, k, regt, gent, pa1t, pa2t,r[1], g[3], pa1[3], pa2[1]]
ll2 = map(str, lll)
print(type(lll))
f.write(str.join(',', ll2) + "\n")
points = points*10
f.close()
#print(s[1])
#print(k_means_regular(datapoints, points, dim, k, allowed_error = 0.001, max_generations = 1)[1])
#print(paper_approach(datapoints, dim, points, k, solution_set_size = 50, top_n = 20, new_n = 0, crossover_points = 2, p = 0.001, allowed_error = 21.5, max_generations = 20, initial_iteration = 5)[1])
#input()
#print(k_means_genetic(datapoints, dim, points, k, solution_set_size = 100, top_n = 10, new_n = 10, crossover_points = 2, p = 1, allowed_error = 21.5, max_generations = 100))
#print(paper_approach_2(datapoints, dim, points, k, solution_set_size = 100, top_n = 20, new_n = 0, crossover_points = 1, p = 0.1, allowed_error = 10, initial_generations = 3, max_iteration = 200)[1])