-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathtw_2g_v2b.py
492 lines (373 loc) · 11.5 KB
/
tw_2g_v2b.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
# -*- coding: utf-8 -*-
"""
Created on Wed May 13 10:00:22 2020
@author: Lenovo
"""
import numpy as np
#from numpy.random import permutation
from scipy.sparse.csgraph import shortest_path
def AccumulateLayerOrder(u0, layerInfo):
# % Input:
# % u0: degree at each layer
# % layerInfo: layer order information
#
# % Output:
# % Accumulate degree at each ordered layers
maxLL = int(max(layerInfo))
u = [0]*(maxLL+1)
for ii in range(maxLL+1):
u[ii] = np.sum(u0[np.where(layerInfo==ii)])
#print(u)
return u
def TW_Chain(u,v):# Tree-Wasserstein for a chain
EPS = 1e-10 # for comparing with 0
n,m = len(u),len(v)
i,j,dd = 0,0,0
while ((i < n) and (j < m)):
if u[i] <= v[j]:
dd = dd + u[i]*np.abs((i/n) - (j/m))
v[j] = v[j] - u[i]
i+=1;
if v[j] <= EPS:
j+=1
else:
dd = dd + v[j]*np.abs((i/n) - (j/m))
u[i] = u[i] - v[j]
j+=1
if u[i] <= EPS:
i+=1
return dd
def TW_InDegrees_NASBENCH(MX, MY,layerX, layerY):
# sum column
u0=np.sum(MX,axis=0)
v0=np.sum(MY,axis=0)
# % Accumulate to vectors using layer_order (also remove isolated nodes)
u = AccumulateLayerOrder(u0, layerX);
v = AccumulateLayerOrder(v0, layerY);
#u=u/np.sum(u)
#v=v/np.sum(v)
if np.sum(u)==0:
u=np.ones(len(u))/len(u)
#print(u)
else:
u=u/np.sum(u)
if np.sum(v)==0:
v=np.ones(len(v))/len(v)
#print(v)
else:
v=v/np.sum(v)
#print(layerX,layerY)
#print(u0,v0)
#print(u,v)
dd = TW_Chain(u, v);
#print("dd",dd)
return dd
def TW_OutDegrees_NASBENCH(MX, MY,layerX, layerY):
# sum row
u0=np.sum(MX,axis=1)
v0=np.sum(MY,axis=1)
# % Accumulate to vectors using layer_order (also remove isolated nodes)
u = AccumulateLayerOrder(u0, layerX);
v = AccumulateLayerOrder(v0, layerY);
# u=u/np.sum(u)
# v=v/np.sum(v)
if np.sum(u)==0:
u=np.ones(len(u))/len(u)
#print(u)
else:
u=u/np.sum(u)
if np.sum(v)==0:
v=np.ones(len(v))/len(v)
#print(v)
else:
v=v/np.sum(v)
dd=TW_Chain(u,v)
return dd
def TreeMetric_Mapping_2G_NASBENCH101(uu):
vv = [0]*21
vv[:12] = uu
# 1-gram subtree
# -- conv
vv[12] = uu[0] + uu[1]
# 1-gram
vv[13] = np.sum(uu[:2])
# 2-gram subtree
# ---- s-conv
vv[14] = uu[3] + uu[7]
# ---- d-conv
vv[15] = uu[4] + uu[6]
# -- conv-conv
vv[16] = uu[3] + uu[7] + uu[4] + uu[6]
# ---- conv-pool
vv[17] = uu[5] + uu[8]
# ---- pool-conv
vv[18] = uu[9] + uu[10]
# -- mix-conv&pool
vv[19] = uu[5] + uu[8] + uu[9] + uu[10]
# 2-gram
vv[20] = np.sum(uu[3:11])
return np.asarray(vv)
def TreeMetric_Mapping_2G_NASBENCH201(uu):
# % Input:
# % uu: 2-gram count vector of operations
# % 2grams count vector
#
# % ---NOTE---
# % NASBENCH201
# % for 4 operations
# % conv1
# % conv3
# % avg_pool
# % sc
#
# % 2grams count vector
# % dim1: #conv1
# % dim2: #conv3
# % dim3: #avg_pool
# % dim4: #sc
#
# % dim5: #conv1-conv1
# % dim6: #conv1-conv3
# % dim7: #conv1-avg_pool
# % dim8: #conv1-sc
#
# % dim9: #conv3-conv1
# % dim10: #conv3-conv3
# % dim11: #conv3-avg_pool
# % dim12: #conv3-sc
#
# % dim13: #avg_pool-conv1
# % dim14: #avg_pool-conv3
# % dim15: #avg_pool-avg-pool
# % dim16: #avg_pool-sc
#
# % dim17: #sc-conv1
# % dim18: #sc-conv3
# % dim19: #sc-avg-pool
# % dim20: #sc-sc
#
# % Input
# % u, v: 2-grams count vector (dim x 1) where dim = 20 for NASBENCH201
# % e.g., u = [#conv1; #conv3; #avg_pool; #sc; ...
# % #conv1-conv1; #conv1-conv3; #conv1-avg_pool; #conv1-sc; ...
# % #conv3-conv1; #conv3-conv3; #conv3-avg_pool; #conv3-sc; ...
# % #avg_pool-conv1; #avg_pool-conv3; #avg_pool-avg_pool; #avg_pool-sc; ...
# % #sc-conv1; #sc-conv3; #sc-avg_pool; #sc-sc]
# % note: using the above format for count vector
vv = [0]*33
vv[0:20] = uu;
# % 1-gram subtree
# % -- conv
vv[20] = uu[1] + uu[2];
# % 1-gram
vv[21] = sum(uu[:3]);
# % 2-gram subtree
# % ---- s-conv
vv[22] = uu[4] + uu[9]
# % ---- d-conv
vv[23] = uu[5] + uu[8]
# % -- conv-conv
vv[24] = uu[4] + uu[9] + uu[5] + uu[8]
# % ---- conv-pool
vv[25] = uu[6] + uu[10]
# % ---- pool-conv
vv[26] = uu[12] + uu[13]
# % -- mix-conv&pool
vv[27] = uu[6] + uu[10] + uu[12] + uu[13]
# % ---- conv-sc
vv[28] = uu[7] + uu[11]
# % ---- sc-conv
vv[29] = uu[16] + uu[17]
# % -- mix-conv&sc
vv[30] = uu[7] + uu[11] + uu[16] + uu[17]
# % -- mix-sc&pool
vv[31] = uu[18] + uu[15]
# % 2-gram
vv[32] = sum(uu[4:19]);
return np.asarray(vv)
def GetTreeMetric_2Grams_Operations_NASBENCH101(alpha):
ww = [alpha, alpha, 1, alpha*alpha, alpha*alpha,
alpha*alpha, alpha*alpha, alpha*alpha, alpha*alpha, alpha*alpha,
alpha*alpha, 1, (1-alpha), 1, alpha*(1-alpha),
alpha*(1-alpha), (1-alpha), alpha*(1-alpha), alpha*(1-alpha), (1-alpha),
1]
return ww
def TW_Operations_v2_NASBENCH101(u,v,alpha=0.1):
# normalisation
u=u/np.sum(u)
v=v/np.sum(v)
"""
if np.sum(u) == 0:
u = np.ones(len(u))/len(u);
else:
u = u/np.sum(u);
#vv=vv/np.sum(vv)
if np.sum(v) == 0:
v = np.ones(len(v))/len(v);
else:
v = v/np.sum(v);
"""
# tree metric mapping
u_tm = TreeMetric_Mapping_2G_NASBENCH101(u)
v_tm = TreeMetric_Mapping_2G_NASBENCH101(v)
#ww_tm = np.asarray([alpha, alpha, 1, 1-alpha])
#ww_tm=GetTreeMetric_2Grams_Operations_NASBENCH101(alpha)
ww_tm = [alpha, alpha, 1, alpha*alpha, alpha*alpha,
alpha*alpha, alpha*alpha, alpha*alpha, alpha*alpha, alpha*alpha,
alpha*alpha, 1, (1-alpha), 1, alpha*(1-alpha),
alpha*(1-alpha), (1-alpha), alpha*(1-alpha), alpha*(1-alpha), (1-alpha),
1]
# tree-Wasserstein
dd = np.sum(ww_tm * np.abs(u_tm - v_tm))
return dd
def TW_Operations_NASBENCH101(u,v,alpha=0.1):
# normalisation
u=u/np.sum(u)
v=v/np.sum(v)
"""
if np.sum(u) == 0:
u = np.ones(len(u))/len(u);
else:
u = u/np.sum(u);
#vv=vv/np.sum(vv)
if np.sum(v) == 0:
v = np.ones(len(v))/len(v);
else:
v = v/np.sum(v);
"""
# tree metric mapping
u_tm = TreeMetric_Mapping_2G_NASBENCH101(u)
v_tm = TreeMetric_Mapping_2G_NASBENCH101(v)
#ww_tm = np.asarray([alpha, alpha, 1, 1-alpha])
#ww_tm=GetTreeMetric_2Grams_Operations_NASBENCH101(alpha)
ww_tm = [alpha, alpha, 1, alpha*alpha, alpha*alpha,
alpha*alpha, alpha*alpha, alpha*alpha, alpha*alpha, alpha*alpha,
alpha*alpha, 1, (1-alpha), 1, alpha*(1-alpha),
alpha*(1-alpha), (1-alpha), alpha*(1-alpha), alpha*(1-alpha), (1-alpha),
1]
# tree-Wasserstein
dd = np.sum(ww_tm * np.abs(u_tm - v_tm))
return dd
def TW_Operations_NB101(u,v,alpha=0.1):
# normalisation
u=u/np.sum(u)
v=v/np.sum(v)
# tree metric mapping
u_tm = u.tolist() + [u[0] + u[1]]
v_tm = v.tolist() +[ v[0] + v[1]]
ww_tm = np.asarray([alpha, alpha, 1, 1-alpha])
u_tm=np.asarray(u_tm)
v_tm=np.asarray(v_tm)
# tree-Wasserstein
dd = np.sum(ww_tm * np.abs(u_tm - v_tm))
return dd
def TW_Operations_NB201(u,v,alpha=0.1):
# can handle both 1Gram and 2Gram
MaxLength=20 # len for 2gram count vector given 4 operations
if len(u)<MaxLength:
u=u+[0]*(MaxLength-len(u))
if len(v)<MaxLength:
v=v+[0]*(MaxLength-len(v))
# normalisation
u=u/np.sum(u)
v=v/np.sum(v)
# tree metric mapping
# u_tm = u.tolist() + [u[0] + u[1]]
# v_tm = v.tolist() +[ v[0] + v[1]]
u_tm = TreeMetric_Mapping_2G_NASBENCH201(u);
v_tm = TreeMetric_Mapping_2G_NASBENCH201(v);
#ww_tm = np.asarray([alpha, alpha, 1, 1-alpha])
alpha = 0.1;
a = alpha;
b = 1-alpha;
aa = a*a;
ab = a*b;
ww_tm = [a ,a ,1 ,1 ,aa , \
aa ,aa ,aa ,aa ,aa , \
aa ,aa ,aa ,aa ,1 , \
a ,aa ,aa ,a ,1 , \
b ,1 ,ab ,ab ,b , \
ab ,ab ,b ,ab ,ab , \
b ,b ,1]
u_tm=np.asarray(u_tm)
v_tm=np.asarray(v_tm)
# tree-Wasserstein
dd = np.sum(ww_tm * np.abs(u_tm - v_tm))
return dd
def TW_NASBENCH101(MX, MY, opX, opY, layerX,layerY):
dd_Operations = TW_Operations_NB101(opX, opY)
#dd_Degrees = TW_Degrees_NASBENCH(MX, MY)
dd_InDegrees = TW_InDegrees_NASBENCH(MX, MY,layerX,layerY)
dd_OutDegrees = TW_OutDegrees_NASBENCH(MX, MY,layerX,layerY)
#dd=lamb*dd_Operations+(1-lamb)*dd_Degrees
return dd_Operations,dd_InDegrees,dd_OutDegrees
def TW_NASBENCH201(MX, MY, opX, opY, layerX,layerY):
dd_Operations = TW_Operations_NB201(opX, opY)
#dd_Degrees = TW_Degrees_NASBENCH(MX, MY)
dd_InDegrees = TW_InDegrees_NASBENCH(MX, MY,layerX,layerY)
dd_OutDegrees = TW_OutDegrees_NASBENCH(MX, MY,layerX,layerY)
#dd=lamb*dd_Operations+(1-lamb)*dd_Degrees
return dd_Operations,dd_InDegrees,dd_OutDegrees
def TW_2G_NB101(MX, MY, opX, opY, layerX,layerY):
dd_Operations = TW_Operations_NASBENCH101(opX, opY)
#dd_Degrees = TW_Degrees_v2_NASBENCH101(MX, MY)
dd_InDegrees = TW_InDegrees_NASBENCH(MX, MY,layerX,layerY)
dd_OutDegrees = TW_OutDegrees_NASBENCH(MX, MY,layerX,layerY)
#print("dd operation",dd_Operations)
#print("dd degree",dd_Degrees)
#dd=lamb*dd_Operations+(1-lamb)*dd_Degrees
#dd=(1-lamb1-lamb2)*dd_Operations+lamb1*dd_InDegrees+lamb2*dd_OutDegrees
#return dd
return dd_Operations, dd_InDegrees,dd_OutDegrees
def TW_2G_NB201(MX, MY, opX, opY,layerX,layerY):
dd_Operations = TW_Operations_NB201(opX, opY)
#dd_Degrees = TW_Degrees_v2_NASBENCH101(MX, MY)
dd_InDegrees = TW_InDegrees_NASBENCH(MX, MY,layerX,layerY)
dd_OutDegrees = TW_OutDegrees_NASBENCH(MX, MY,layerX,layerY)
#print("dd operation",dd_Operations)
#print("dd degree",dd_Degrees)
#dd=lamb*dd_Operations+(1-lamb)*dd_Degrees
#dd=(1-lamb1-lamb2)*dd_Operations+lamb1*dd_InDegrees+lamb2*dd_OutDegrees
#return dd
return dd_Operations, dd_InDegrees,dd_OutDegrees
MZ = np.asarray([[0, 1, 1, 0, 1, 0, 0],
[0,0,0,0,1,1, 1],
[0,0,0,0,0,0, 0],
[0,0,0,0, 0,0,0],
[0,0,0,0,0,0, 0],
[0,0,0,0,0,0, 0],
[0,0,0,0,0,0,0]])
MX = np.asarray([[0, 1, 1, 1, 0, 1, 0],
[0,0,0,0,0,0, 1],
[0,0,0,0,0,0, 1],
[0,0,0,0, 1,0,0],
[0,0,0,0,0,0, 1],
[0,0,0,0,0,0, 1],
[0,0,0,0,0,0,0]])
#% adjacency matrix for network B
MY = np.asarray( [[0, 1, 0, 1, 0, 1, 0],
[ 1,0,0,1,1,0, 1],
[0,0,1,0,0,0, 1],
[0,0,0,0, 1,0,0],
[0,0,0,0,0,0, 1],
[0,0,1,0,0,0, 1],
[0,0,0,0,0,0,0]])
opX=[4,3,2,1,1,2,3,1,4,2,0,1]
opY=[4,3,2,1,1,2,3,1,3,2,0,1]
#dd=TW_v2_NASBENCH101(MX,MY,opX,opY)
##print(dd)
#
#dd=TW_v2_NASBENCH101(MZ,MZ,opX,opX)
#print(dd)
layerX=shortest_path(MX,method="D")
layerX[layerX==np.inf]=-1
layerX=layerX[0,:]
layerY=shortest_path(MY,method="D")
layerY[layerY==np.inf]=-1
layerY=layerY[0,:]
dd=TW_2G_NB101(MX,MY,opX,opY,layerX,layerY)
#print(dd)
dd=TW_2G_NB101(MZ,MZ,opX,opX,layerX,layerY)
#print(dd)
dd=TW_2G_NB201(MX,MZ,opX,opX,layerX,layerY)
#print(dd)