-
Notifications
You must be signed in to change notification settings - Fork 3
/
hyperparam_gridsearch.py
332 lines (281 loc) · 13.3 KB
/
hyperparam_gridsearch.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
"""Iterate through hyperparams of ST to get accuracies
"""
import os
import time
import numpy as np
from sklearn.svm import SVC
from tqdm import tqdm
import hyper_pixelNN as hsinn
import pdb
FEAT_DIR = '/scratch0/ilya/locDoc/data/hyperspec/features/npz_feat'
FEAT_DIR = '/scratch1/ilya/locDoc/data/hyperspec/features/npz_feat'
FEAT_DIR = '/scratch2/ilyak/locDoc/data/hyperspec/features/npz_feat'
RESULTS_DIR = '/scratch0/ilya/locDoc/pyfst'
RESULTS_DIR = '/scratch2/ilyak/locDoc/pyfst'
def run_svm(trainX, trainY, testX, testY):
print('starting training')
start = time.time()
clf = SVC(kernel='linear', C=1000.0)
clf.fit(trainX, trainY)
end = time.time()
print('Training done. Took %is' % int(end - start))
# now test
test_chunk_size = 1000
# acc = clf.score(testX, testY)
# better to score in batches to limit memory use:
n_correct = 0
for i in tqdm(range(0,testY.shape[0],test_chunk_size), desc='Testing'):
p_label = clf.predict(testX[i:i+test_chunk_size]);
n_correct += (p_label == testY[i:i+test_chunk_size]).sum()
acc = float(n_correct) / testY.shape[0]
return acc
def get_acc_for_config(dataset, st_net_spec, preprocessed_data_path, preprocessed_data_root, mask_paths, save_features):
"""One set of hyperparams and dataset, several trials
"""
trainimgname, trainlabelname = hsinn.dset_filenames_dict[dataset]
trainimgfield, trainlabelfield = hsinn.dset_fieldnames_dict[dataset]
data, labels = hsinn.load_data(trainimgname, trainimgfield, trainlabelname, trainlabelfield)
if save_features:
data = hsinn.load_or_preprocess_data(data, preprocessed_data_path, preprocessed_data_root, st_net_spec, 31)
else:
data = hsinn.preprocess_data(data, st_net_spec, 31)
accs = []
# loop through masks
for mask_path in mask_paths:
train_mask = hsinn.multiversion_matfile_get_field(mask_path, 'train_mask')
val_mask = hsinn.multiversion_matfile_get_field(mask_path, 'test_mask')
trainX, trainY, valX, valY = hsinn.get_train_val_splits(data, labels, train_mask, val_mask, (0,0,0)) # , n_eval=2048
# run svm
accs.append(run_svm(trainX.squeeze(), trainY, valX.squeeze(), valY))
print('[%s] Avg acc is %.2f' % (hsinn.spec_to_str(st_net_spec), sum(accs) / len(accs)))
return accs
def specs1():
"""
Spatial psis should be decreasing, phi should be equal to smallest psi
"""
specs = []
myrange = [9,7,5,3]
for b in myrange:
for s1 in myrange:
for s2 in myrange:
if s1 >= s2:
specs.append( [ [b,s1,s1], [b,s2,s2], [b,s2,s2] ] )
return specs
def specs2():
"""
Spatial psis should be decreasing, phi should be equal to smallest psi
"""
specs = []
myrange = [7,5,3,1]
for b in [9,7,5,3]:
for s1 in myrange:
for s2 in myrange:
if s1 >= s2:
specs.append( [ [b,s1,s1], [b,s2,s2], [b,s2,s2] ] )
return specs
def specs_append1():
"""
Spatial psis should be decreasing, phi should be equal to smallest psi
"""
specs = []
myrange = [9,7,5,3]
for b in myrange:
for s1 in [1]:
for s2 in [1]:
if s1 >= s2:
specs.append( [ [b,s1,s1], [b,s2,s2], [b,s2,s2] ] )
return specs
def specs_append9():
"""
Spatial psis should be decreasing, phi should be equal to smallest psi
"""
specs = []
myrange = [9,7,5,3]
for b in myrange:
for s1 in [9]:
for s2 in myrange:
if s1 >= s2:
specs.append( [ [b,s1,s1], [b,s2,s2], [b,s2,s2] ] )
return specs
def specs_append9_bands():
"""
Spatial psis should be decreasing, phi should be equal to smallest psi
"""
specs = []
myrange = [7,5,3]
for b in [9]:
for s1 in myrange:
for s2 in myrange:
if s1 >= s2:
specs.append( [ [b,s1,s1], [b,s2,s2], [b,s2,s2] ] )
return specs
def perform_gridsearch(dataset, masks, specs, outfile='gridsearch.npz', save_features=False, preprocessed_data_root=FEAT_DIR):
# a list< list < int > > -> list< float >.
# Maps a configuration to a set of accuracies
results = {}
for spec_i, spec in enumerate(specs):
psi1,psi2,phi = spec
st_net_spec = hsinn.st_net_spec_struct(psi1,psi2,phi)
preprocessed_data_path = os.path.join(preprocessed_data_root, '%s__%s.npz' % (dataset, hsinn.spec_to_str(st_net_spec)))
accs = get_acc_for_config(dataset, st_net_spec, preprocessed_data_path, preprocessed_data_root, masks, save_features)
results[hsinn.spec_to_str(st_net_spec)] = accs
print('FINISHED %i/%i' % (spec_i+1, len(specs)))
np.savez(os.path.join(RESULTS_DIR, outfile), results=results)
print('Saved %s' % os.path.join(RESULTS_DIR, outfile))
def paviaU():
dataset = 'PaviaU'
# masks = ['PaviaU_gt_traintest_s03_1_3f6384.mat',
# 'PaviaU_gt_traintest_s03_2_b67b5f.mat',
# 'PaviaU_gt_traintest_s03_3_7d8356.mat',
# 'PaviaU_gt_traintest_s03_4_241266.mat',
# 'PaviaU_gt_traintest_s03_5_ccbbb1.mat',
# 'PaviaU_gt_traintest_s03_6_dce186.mat',
# 'PaviaU_gt_traintest_s03_7_d5cdfe.mat',
# 'PaviaU_gt_traintest_s03_8_6bcd5a.mat',
# 'PaviaU_gt_traintest_s03_9_a1ff2b.mat',
# 'PaviaU_gt_traintest_s03_10_e1dac2.mat']
# masks = ['PaviaU_strictsinglesite_trainval_s90_0_700083.mat',
# 'PaviaU_strictsinglesite_trainval_s90_1_509297.mat',
# 'PaviaU_strictsinglesite_trainval_s90_2_291093.mat',
# 'PaviaU_strictsinglesite_trainval_s90_3_232341.mat',
# 'PaviaU_strictsinglesite_trainval_s90_4_833337.mat',
# 'PaviaU_strictsinglesite_trainval_s90_5_672053.mat',
# 'PaviaU_strictsinglesite_trainval_s90_6_215589.mat',
# 'PaviaU_strictsinglesite_trainval_s90_7_553699.mat',
# 'PaviaU_strictsinglesite_trainval_s90_8_734449.mat',
# 'PaviaU_strictsinglesite_trainval_s90_9_484519.mat']
masks = ['PaviaU_distributed_trainval_p0200_0_112681.mat',
'PaviaU_distributed_trainval_p0200_1_678193.mat',
'PaviaU_distributed_trainval_p0200_2_123909.mat',
'PaviaU_distributed_trainval_p0200_3_346269.mat',
'PaviaU_distributed_trainval_p0200_4_987781.mat',
'PaviaU_distributed_trainval_p0200_5_057029.mat',
'PaviaU_distributed_trainval_p0200_6_925171.mat',
'PaviaU_distributed_trainval_p0200_7_971587.mat',
'PaviaU_distributed_trainval_p0200_8_776217.mat',
'PaviaU_distributed_trainval_p0200_9_216285.mat']
# masks = [os.path.join('/scratch0/ilya/locDoc/data/hyperspec/',m) for m in masks]
# perform_gridsearch(dataset, masks, specs1(), outfile='gridsearch_paviaU_s200_10trials_full_new.npz', save_features=True)
# masks = ['PaviaU_strictsinglesite_trainval_s50_0_979087.mat',
# 'PaviaU_strictsinglesite_trainval_s50_1_270161.mat',
# 'PaviaU_strictsinglesite_trainval_s50_2_259715.mat',
# 'PaviaU_strictsinglesite_trainval_s50_3_107251.mat',
# 'PaviaU_strictsinglesite_trainval_s50_4_729473.mat',
# 'PaviaU_strictsinglesite_trainval_s50_5_129325.mat',
# 'PaviaU_strictsinglesite_trainval_s50_6_927653.mat',
# 'PaviaU_strictsinglesite_trainval_s50_7_627051.mat',
# 'PaviaU_strictsinglesite_trainval_s50_8_525881.mat',
# 'PaviaU_strictsinglesite_trainval_s50_9_785489.mat']
masks = [os.path.join('/cfarhomes/ilyak/[email protected]/ramawks69/pyfst/masks/',m) for m in masks]
perform_gridsearch(dataset, masks, specs1(), outfile='gridsearch_june19_paviaU_distributed_p0200_10trials.npz', save_features=True)
def Botswana():
dataset = 'Botswana'
# masks = ['Botswana_strictsinglesite_trainval_s20_0_840407.mat',
# 'Botswana_strictsinglesite_trainval_s20_1_027359.mat',
# 'Botswana_strictsinglesite_trainval_s20_2_059593.mat',
# 'Botswana_strictsinglesite_trainval_s20_3_800757.mat',
# 'Botswana_strictsinglesite_trainval_s20_4_848729.mat',
# 'Botswana_strictsinglesite_trainval_s20_5_309369.mat',
# 'Botswana_strictsinglesite_trainval_s20_6_913005.mat',
# 'Botswana_strictsinglesite_trainval_s20_7_508879.mat',
# 'Botswana_strictsinglesite_trainval_s20_8_218687.mat',
# 'Botswana_strictsinglesite_trainval_s20_9_288573.mat']
# masks = ['Botswana_strictsinglesite_trainval_s20_0_619497.mat',
# 'Botswana_strictsinglesite_trainval_s20_1_371565.mat',
# 'Botswana_strictsinglesite_trainval_s20_2_133689.mat',
# 'Botswana_strictsinglesite_trainval_s20_3_578661.mat',
# 'Botswana_strictsinglesite_trainval_s20_4_954915.mat',
# 'Botswana_strictsinglesite_trainval_s20_5_406769.mat',
# 'Botswana_strictsinglesite_trainval_s20_6_929435.mat',
# 'Botswana_strictsinglesite_trainval_s20_7_053419.mat',
# 'Botswana_strictsinglesite_trainval_s20_8_379701.mat',
# 'Botswana_strictsinglesite_trainval_s20_9_900903.mat']
masks = ['Botswana_distributed_trainval_s20_0_383283.mat',
'Botswana_distributed_trainval_s20_1_935671.mat',
'Botswana_distributed_trainval_s20_2_145243.mat',
'Botswana_distributed_trainval_s20_3_753761.mat',
'Botswana_distributed_trainval_s20_4_716839.mat',
'Botswana_distributed_trainval_s20_5_817361.mat',
'Botswana_distributed_trainval_s20_6_315569.mat',
'Botswana_distributed_trainval_s20_7_737133.mat',
'Botswana_distributed_trainval_s20_8_637855.mat',
'Botswana_distributed_trainval_s20_9_289575.mat']
masks = [os.path.join('/cfarhomes/ilyak/[email protected]/ramawks69/pyfst/masks/',m) for m in masks]
perform_gridsearch(dataset, masks, specs1(), outfile='gridsearch_june19_Bots_distributed_s20_10trials.npz', save_features=True)
def KSC():
dataset = 'KSC'
# masks = ['KSC_strictsinglesite_trainval_s50_0_665848.mat',
# 'KSC_strictsinglesite_trainval_s50_1_783796.mat',
# 'KSC_strictsinglesite_trainval_s50_2_455308.mat',
# 'KSC_strictsinglesite_trainval_s50_3_643924.mat',
# 'KSC_strictsinglesite_trainval_s50_4_514244.mat',
# 'KSC_strictsinglesite_trainval_s50_5_762052.mat',
# 'KSC_strictsinglesite_trainval_s50_6_668196.mat',
# 'KSC_strictsinglesite_trainval_s50_7_990120.mat',
# 'KSC_strictsinglesite_trainval_s50_8_191432.mat',
# 'KSC_strictsinglesite_trainval_s50_9_307364.mat']
masks = ['KSC_distributed_trainval_s20_0_128852.mat',
'KSC_distributed_trainval_s20_1_644908.mat',
'KSC_distributed_trainval_s20_2_182432.mat',
'KSC_distributed_trainval_s20_3_100036.mat',
'KSC_distributed_trainval_s20_4_353740.mat',
'KSC_distributed_trainval_s20_5_348604.mat',
'KSC_distributed_trainval_s20_6_263496.mat',
'KSC_distributed_trainval_s20_7_751704.mat',
'KSC_distributed_trainval_s20_8_029188.mat',
'KSC_distributed_trainval_s20_9_760600.mat']
masks = [os.path.join('/cfarhomes/ilyak/[email protected]/ramawks69/pyfst/masks/', m) for m in masks]
perform_gridsearch(dataset, masks, specs1(), outfile='gridsearch_june19_KSC_distributed_s20_10trials.npz', save_features=True)
def IP():
dataset = 'IP'
# masks = ['IP_strictsinglesite_trainval_s20_0_211328.mat',
# 'IP_strictsinglesite_trainval_s20_1_881136.mat',
# 'IP_strictsinglesite_trainval_s20_2_869530.mat',
# 'IP_strictsinglesite_trainval_s20_3_014114.mat',
# 'IP_strictsinglesite_trainval_s20_4_586540.mat',
# 'IP_strictsinglesite_trainval_s20_5_805378.mat',
# 'IP_strictsinglesite_trainval_s20_6_666022.mat',
# 'IP_strictsinglesite_trainval_s20_7_581782.mat',
# 'IP_strictsinglesite_trainval_s20_8_901868.mat',
# 'IP_strictsinglesite_trainval_s20_9_778852.mat']
# masks = ['IP_strictsinglesite_trainval_p2000_0_533076.mat',
# 'IP_strictsinglesite_trainval_p2000_1_494066.mat',
# 'IP_strictsinglesite_trainval_p2000_2_984650.mat',
# 'IP_strictsinglesite_trainval_p2000_3_340434.mat',
# 'IP_strictsinglesite_trainval_p2000_4_605220.mat',
# 'IP_strictsinglesite_trainval_p2000_5_199324.mat',
# 'IP_strictsinglesite_trainval_p2000_6_596068.mat',
# 'IP_strictsinglesite_trainval_p2000_7_199664.mat',
# 'IP_strictsinglesite_trainval_p2000_8_881256.mat',
# 'IP_strictsinglesite_trainval_p2000_9_005288.mat']
masks = ['IP_distributed_trainval_p1000_0_970730.mat',
'IP_distributed_trainval_p1000_1_532502.mat',
'IP_distributed_trainval_p1000_2_339308.mat',
'IP_distributed_trainval_p1000_3_756872.mat',
'IP_distributed_trainval_p1000_4_633060.mat',
'IP_distributed_trainval_p1000_5_914454.mat',
'IP_distributed_trainval_p1000_6_667722.mat',
'IP_distributed_trainval_p1000_7_240942.mat',
'IP_distributed_trainval_p1000_8_272914.mat',
'IP_distributed_trainval_p1000_9_521992.mat']
masks = [os.path.join('/cfarhomes/ilyak/[email protected]/ramawks69/pyfst/masks/', m) for m in masks]
perform_gridsearch(dataset, masks, specs1(), outfile='gridsearch_june19b_IP_distributed_p10_10trials.npz', save_features=True)
masks = ['IP_strictsinglesite_trainval_p1000_0_766900.mat',
'IP_strictsinglesite_trainval_p1000_1_413194.mat',
'IP_strictsinglesite_trainval_p1000_2_252770.mat',
'IP_strictsinglesite_trainval_p1000_3_582176.mat',
'IP_strictsinglesite_trainval_p1000_4_834150.mat',
'IP_strictsinglesite_trainval_p1000_5_599784.mat',
'IP_strictsinglesite_trainval_p1000_6_076514.mat',
'IP_strictsinglesite_trainval_p1000_7_168940.mat',
'IP_strictsinglesite_trainval_p1000_8_531590.mat',
'IP_strictsinglesite_trainval_p1000_9_613356.mat']
masks = [os.path.join('/cfarhomes/ilyak/[email protected]/ramawks69/pyfst/masks/', m) for m in masks]
perform_gridsearch(dataset, masks, specs2(), outfile='gridsearch_june19_IP_strictsinglesite_p10_10trials.npz', save_features=True)
def main():
# paviaU()
# IP()
KSC()
# Botswana()
if __name__ == '__main__':
main()