forked from maheshakya/Performance_evaluations_ANN
/
evaluation_functions.py
494 lines (366 loc) · 14.9 KB
/
evaluation_functions.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
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
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
import numpy as np
import scipy as sp
import pickle, time
from sklearn.neighbors import NearestNeighbors
from annoy import AnnoyIndex
from pykgraph import KGraph
from pyflann import FLANN
from lshash import LSHash
from lsh_forest import LSH_forest
import nearpy
from nearpy import Engine
from nearpy.hashes import RandomBinaryProjections
#sklearn class
class sklearn_funtions():
def build_index(self, X):
nbrs = NearestNeighbors()
nbrs.fit(X)
return nbrs
def query(self, X, t, query_vector, number_of_neighbors):
distances, indices = t.kneighbors(X[query_vector], n_neighbors=number_of_neighbors)
return list(indices[0])
def query_random_vector(self, t, query_vector, number_of_neighbors):
distances, indices = t.kneighbors(query_vector, n_neighbors=number_of_neighbors)
return list(indices[0])
def query_params(self, X, index_structure, query_vector, number_of_neighbors):
params = {'X':X, 't':index_structure, 'query_vector':query_vector, 'number_of_neighbors':number_of_neighbors}
return params
#ANNOY class
class ANNOY_functions():
def build_index(self,X):
f = X.shape[1]
n = X.shape[0]
t = AnnoyIndex(f, metric = 'angular')
for i in range(n):
t.add_item(i, X[i].tolist())
t.build(-1)
return t
def query(self, t, query_vector, number_of_neighbors):
return t.get_nns_by_item(query_vector, number_of_neighbors)
def query_random_vector(self, t, query_vector, number_of_neighbors):
return t.get_nns_by_vector(query_vector.tolist(), number_of_neighbors)
def query_params(self, X, index_structure, query_vector, number_of_neighbors):
params = {'t':index_structure, 'query_vector':query_vector, 'number_of_neighbors':number_of_neighbors}
return params
#KGraph class
class KGraph_functions():
def build_index(self, X):
index = KGraph()
index.build(X)
return index
def query(self, X, t, query_vector, number_of_neighbors):
return t.search(data = X, query = X[query_vector:query_vector+1], K = number_of_neighbors)[0]
def query_random_vector(self, X, t, query_vector, number_of_neighbors):
return t.search(data = X, query = np.array([query_vector]), K = number_of_neighbors)[0]
def query_params(self, X, index_structure, query_vector, number_of_neighbors):
params = {'X':X, 't':index_structure, 'query_vector':query_vector, 'number_of_neighbors':number_of_neighbors}
return params
#FLANN class
class FLANN_funtions():
def build_index(self, X):
flann = FLANN()
params = flann.build_index(X)
return flann
def query(self, X, t, query_vector, number_of_neighbors):
neighbors, dists = t.nn_index(X[query_vector], number_of_neighbors)
return neighbors[0]
def query_random_vector(self, X, t, query_vector, number_of_neighbors):
neighbors, dists = t.nn_index(query_vector, number_of_neighbors)
return neighbors[0]
def query_params(self, X, index_structure, query_vector, number_of_neighbors):
params = {'X':X, 't':index_structure, 'query_vector':query_vector, 'number_of_neighbors':number_of_neighbors}
return params
#nearpy class
class nearpy_funtions():
def build_index(self, X):
f = X.shape[1]
n = X.shape[0]
rbp = RandomBinaryProjections('rbp', 32)
engine = Engine(f, lshashes=[rbp])
for i in range(n):
engine.store_vector(X[i], 'data_%d' % i)
return engine
def query(sefl, X, t, query_vector):
return t.neighbours(X[query_vector])
def query_random_vector(self, X, t, query_vector):
return t.neighbours(query_vector)
def query_params(self, X, index_structure, query_vector, number_of_neighbors):
params = {'X':X, 't': index_structure, 'query_vector':query_vector}
return params
#lshash class
class lshash_functions():
def build_index(self, X):
f = X.shape[1]
n = X.shape[0]
lsh = LSHash(hash_size = 32, input_dim = f, num_hashtables = 100)
for i in range(n):
lsh.index(X[i], i)
return lsh
def query(self, X, t, query_vector):
return t.query(query_point = X[query_vector])
def query_random_vector(self, X, t, query_vector):
return t.query(query_point = query_vector)
def query_params(self, X, index_structure, query_vector, number_of_neighbors):
params = {'X':X, 't': index_structure, 'query_vector':query_vector}
return params
#lsh_forest class
class lsh_forest_functions():
def build_index(self, X):
f = X.shape[1]
n = X.shape[0]
lsh_forest = LSH_forest(number_of_trees=10)
lsh_forest.build_index(X)
return lsh_forest
def query(self, X, t, query_vector, c):
return t.query(X[query_vector], c=c)
def query_random_vector(self, X, t, query_vector, c):
return t.query(query_vector, c=c)
def query_params(self, X, index_structure, query_vector, c):
params = {'X':X, 't':index_structure, 'query_vector':query_vector, 'c':c}
return params
"""
Query speed tests for nearest neighbor
"""
#Qyery speed tests for nearest neighbor : data point in the data set
def queryTimes_nearest(X, imp = ANNOY_functions):
f = X.shape[1]
n = X.shape[0]
ann = imp()
t = ann.build_index(X)
lim_n = 1000
time_sum = 0
for i in xrange(lim_n):
j = np.random.randint(0, n)
print 'finding the nearest neighnor for', j
params = ann.query_params(X, t, j, 1)
t0 = time.time()
neighbors = ann.query(**params)
T = time.time() - t0
time_sum = time_sum + T
print 'avg time: %.6fs' % (time_sum / (i + 1))
return time_sum / lim_n
def queryTimes_nearest_LSHF(X, imp = lsh_forest_functions, c=1):
f = X.shape[1]
n = X.shape[0]
ann = imp()
t = ann.build_index(X)
lim_n = 1000
time_sum = 0
for i in xrange(lim_n):
j = np.random.randint(0, n)
print 'finding the nearest neighnor for', j
params = ann.query_params(X, t, j, c)
t0 = time.time()
neighbors = ann.query(**params)
T = time.time() - t0
time_sum = time_sum + T
print 'avg time: %.6fs' % (time_sum / (i + 1))
return time_sum / lim_n
#Qyery speed tests for nearest neighbor : a random data point
def queryTimes_nearest_random_vector(X, imp = ANNOY_functions):
f = X.shape[1]
n = X.shape[0]
ann = imp()
t = ann.build_index(X)
lim_n = 1000
time_sum = 0
for i in xrange(lim_n):
query_vector = np.random.random(size=f)
print 'finding the nearest neighnor for a random vector'
params = ann.query_params(X, t, query_vector, 1)
t0 = time.time()
neighbors = ann.query(**params)
T = time.time() - t0
time_sum = time_sum + T
print 'avg time: %.6fs' % (time_sum / (i + 1))
return time_sum / lim_n
def queryTimes_nearest_random_vector_LSHF(X, imp = lsh_forest_functions, c=1):
f = X.shape[1]
n = X.shape[0]
ann = imp()
t = ann.build_index(X)
lim_n = 1000
time_sum = 0
for i in xrange(lim_n):
query_vector = np.random.random(size=f)
print 'finding the nearest neighnor for a random vector'
params = ann.query_params(X, t, query_vector, c)
t0 = time.time()
neighbors = ann.query(**params)
T = time.time() - t0
time_sum = time_sum + T
print 'avg time: %.6fs' % (time_sum / (i + 1))
return time_sum / lim_n
"""
Query speed tests for multiple nearest neighbors
"""
# a data point in the data set
def queryTimes_multiple(X, limits = [1, 10, 100, 1000] , imp = ANNOY_functions, iterations = 100):
f = X.shape[1]
n = X.shape[0]
ann = imp()
t = ann.build_index(X)
limits = limits
lim_n = iterations
time_sum = {}
for i in xrange(lim_n):
j = np.random.randint(0, n)
print 'finding nbs for', j
for limit in limits:
params = ann.query_params(X, t, j, limit)
t0 = time.time()
neighbors = ann.query(**params)
T = time.time() - t0
time_sum[limit] = time_sum.get(limit, 0.0) + T
for limit in limits:
print 'limit: %-9d avg time: %.6fs' % (limit, time_sum[limit] / (i + 1))
return time_sum
# a random data point
def queryTimes_multiple_random_vector(X, limits = [1, 10, 100, 1000] , imp = ANNOY_functions, iterations = 100):
f = X.shape[1]
n = X.shape[0]
ann = imp()
t = ann.build_index(X)
limits = limits
lim_n = iterations
time_sum = {}
for i in xrange(lim_n):
query_vector = np.random.random(size=f)
print 'finding the nearest neighnor for a random vector'
for limit in limits:
params = ann.query_params(X, t, query_vector, limit)
t0 = time.time()
neighbors = ann.query(**params)
T = time.time() - t0
time_sum[limit] = time_sum.get(limit, 0.0) + T
for limit in limits:
print 'limit: %-9d avg time: %.6fs' % (limit, time_sum[limit] / (i + 1))
return time_sum
"""
Precision tests for ANNOY, FLANN, KGraph
"""
# a data point in the data set
def precision_test(X, limits = [10, 100, 1000] , imp = ANNOY_functions, iterations = 100):
f = X.shape[1]
n = X.shape[0]
ann = imp()
t = ann.build_index(X)
limits = limits
k = 10
prec_sum = {}
prec_n = iterations
time_sum = {}
for i in xrange(prec_n):
j = np.random.randint(0, n)
print 'finding nbs for', j
params = ann.query_params(X, t, j, n)
closest = set(ann.query(**params)[:k])
for limit in limits:
params = ann.query_params(X, t, j, limit)
t0 = time.time()
toplist = ann.query(**params)
T = time.time() - t0
found = len(closest.intersection(toplist))
hitrate = 1.0 * found / k
prec_sum[limit] = prec_sum.get(limit, 0.0) + hitrate
time_sum[limit] = time_sum.get(limit, 0.0) + T
for limit in limits:
print 'limit: %-9d precision: %6.2f%% avg time: %.6fs' % (limit, 100.0 *
prec_sum[limit] / (i + 1), time_sum[limit] / (i + 1))
return prec_sum, time_sum
# a random data point
def precision_test_random_vector(X, limits = [10, 100, 1000] , imp = ANNOY_functions, iterations = 100):
f = X.shape[1]
n = X.shape[0]
ann = imp()
t = ann.build_index(X)
limits = limits
k = 10
prec_sum = {}
prec_n = iterations
time_sum = {}
for i in xrange(prec_n):
query_vector = np.random.random(size=f)
print 'finding the nearest neighnor for a random vector'
params = ann.query_params(X, t, query_vector, n)
closest = set(ann.query_random_vector(**params)[:k])
for limit in limits:
params = ann.query_params(X, t, query_vector, limit)
t0 = time.time()
toplist = ann.query_random_vector(**params)
T = time.time() - t0
found = len(closest.intersection(toplist))
hitrate = 1.0 * found / k
prec_sum[limit] = prec_sum.get(limit, 0.0) + hitrate
time_sum[limit] = time_sum.get(limit, 0.0) + T
for limit in limits:
print 'limit: %-9d precision: %6.2f%% avg time: %.6fs' % (limit, 100.0 *
prec_sum[limit] / (i + 1), time_sum[limit] / (i + 1))
return prec_sum, time_sum
"""
Precision test for lsh_forest
"""
# a data point in the data set
def precision_test_candidates_LSH_F(X, limits = [10, 100, 1000], iterations = 100):
f = X.shape[1]
n = X.shape[0]
t = LSH_forest(number_of_trees=10)
t.build_index(X)
limits = limits
k = 10
prec_sum = {}
prec_n = iterations
time_sum = {}
candidate_sum = {}
for i in xrange(prec_n):
j = np.random.randint(0, n)
print 'finding nbs for', j
neighbors, candidates = t.query_num_candidates(X[j], c=n)
closest = set(neighbors[:k])
for limit in limits:
t0 = time.time()
neighbors, candidates= t.query_num_candidates(X[j], c=limit)
T = time.time() - t0
toplist = neighbors[:k]
found = len(closest.intersection(toplist))
hitrate = 1.0 * found / k
prec_sum[limit] = prec_sum.get(limit, 0.0) + hitrate
time_sum[limit] = time_sum.get(limit, 0.0) + T
candidate_sum[limit] = candidate_sum.get(limit, 0) + candidates
for limit in limits:
print 'limit: %.8f precision: %6.2f%% avg time: %.6fs avg_candiates: %.1f' % (limit, 100.0 *
prec_sum[limit] / (i + 1), time_sum[limit] / (i + 1),
candidate_sum[limit]/ (i+1))
return prec_sum, time_sum, candidate_sum
# random vector
def precision_test_candidates_LSH_F_random(X, limits = [10, 100, 1000], iterations = 100):
f = X.shape[1]
n = X.shape[0]
t = LSH_forest(number_of_trees=10)
t.build_index(X)
limits = limits
k = 10
prec_sum = {}
prec_n = iterations
time_sum = {}
candidate_sum = {}
for i in xrange(prec_n):
query_vector = np.random.random(size=f)
print 'finding nbs for a random vector'
neighbors, candidates = t.query_num_candidates(query_vector, c=n)
closest = set(neighbors[:k])
for limit in limits:
t0 = time.time()
neighbors, candidates= t.query_num_candidates(query_vector, c=limit)
T = time.time() - t0
toplist = neighbors[:k]
found = len(closest.intersection(toplist))
hitrate = 1.0 * found / k
prec_sum[limit] = prec_sum.get(limit, 0.0) + hitrate
time_sum[limit] = time_sum.get(limit, 0.0) + T
candidate_sum[limit] = candidate_sum.get(limit, 0) + candidates
for limit in limits:
print 'limit: %.8f precision: %6.2f%% avg time: %.6fs avg_candiates: %.1f' % (limit, 100.0 *
prec_sum[limit] / (i + 1), time_sum[limit] / (i + 1),
candidate_sum[limit]/ (i+1))
return prec_sum, time_sum, candidate_sum