-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathbuilding_detection_large.py
230 lines (191 loc) · 8.59 KB
/
building_detection_large.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
import sys
from pathlib import Path
import dill
import numpy as np
from sklearn.svm import SVC
from qsvm import QSVM
from quantum_kernels import (
Kernel,
data_reuploading_feature_map,
)
from utils import accuracy, confusion_matrix, matthews_corrcoef, f1_score
def run_large():
n_train_samples = 10_000
n_valid_samples = 100_000
with open(DATA_DIR / f'dataset_{DATASET}_ts={n_train_samples}_vs={n_valid_samples}.pkl', 'rb') as f:
train_x, train_y, valid_x, valid_y = dill.load(f)
n_train_samples = 5_000
train_x = train_x[:n_train_samples]
train_y = train_y[:n_train_samples]
train_mean = np.mean(train_x, axis=0)
train_std = np.std(train_x, axis=0)
train_x_normalized = (train_x - train_mean) / train_std
valid_x_normalized = (valid_x - train_mean) / train_std
###################################################################################################################
# SVM #############################################################################################################
###################################################################################################################
svm_common_params = dict(kernel='rbf', class_weight='balanced')
if DATASET == 'kits':
params = dict(C=0.1778279410038923, gamma=1.0)
elif DATASET == 'downtown':
params = dict(C=5.623413251903491, gamma=1.0)
elif DATASET == 'ptgrey':
params = dict(C=0.1, gamma=5.623413251903491)
params |= svm_common_params
svm = SVC(**params)
svm.fit(train_x_normalized, train_y)
with open(LOG_DIR / f'svm_{DATASET}_ts={n_train_samples}.pkl', 'wb') as f:
dill.dump(svm, f)
with open(LOG_DIR / f'svm_{DATASET}_ts={n_train_samples}.pkl', 'rb') as f:
svm = dill.load(f)
preds = svm.predict(valid_x_normalized)
cm = confusion_matrix(preds, valid_y)
mcc = matthews_corrcoef(cm)
f1 = f1_score(cm)
acc = accuracy(cm)
print(f'SVM:\n\tcm = {cm.tolist()}\n\t{mcc = :.3f}\n\t{f1 = :.3f}\n\t{acc = :.2%}')
del svm
###################################################################################################################
# QSVM ############################################################################################################
###################################################################################################################
qsvm_sampler = 'hybrid'
qsvm_hybrid_time_limit = 240
qsvm_common_params = dict(
kernel='rbf',
sampler=qsvm_sampler,
hybrid_time_limit=qsvm_hybrid_time_limit,
num_reads=1_000,
normalize=True,
optimize_memory=False,
dwave_api_token=None,
fail_to_classical=False,
)
if DATASET == 'kits':
params = dict(
B=2, P=1, K=4, zeta=0.0, gamma=0.1778279410038923, threshold=0.0001, threshold_strategy='relative'
)
elif DATASET == 'downtown':
params = dict(B=2, P=2, K=4, zeta=0.4, gamma=1.0, threshold=0.0001, threshold_strategy='relative')
elif DATASET == 'ptgrey':
params = dict(
B=2, P=2, K=4, zeta=1.2, gamma=1.7782794100389228, threshold=0.0001, threshold_strategy='relative'
)
params |= qsvm_common_params
qsvm = QSVM(**params)
qsvm.fit(train_x, train_y)
with open(
LOG_DIR / f'qsvm_2_{DATASET}_ts={n_train_samples}_sampler={qsvm_sampler}_htl={qsvm_hybrid_time_limit}.pkl',
'wb',
) as f:
dill.dump(qsvm, f)
with open(
LOG_DIR / f'qsvm_{DATASET}_ts={n_train_samples}_sampler={qsvm_sampler}_htl={qsvm_hybrid_time_limit}.pkl',
'rb',
) as f:
qsvm = dill.load(f)
preds = qsvm.predict(valid_x)
cm = confusion_matrix(preds, valid_y)
mcc = matthews_corrcoef(cm)
f1 = f1_score(cm)
acc = accuracy(cm)
print(f'QSVM:\n\tcm = {cm.tolist()}\n\t{mcc = :.3f}\n\t{f1 = :.3f}\n\t{acc = :.2%}')
del qsvm
###################################################################################################################
# SVM w/ Quantum Kernel ###########################################################################################
###################################################################################################################
kernel_svm_common_params = dict(class_weight='balanced')
if DATASET == 'kits':
kernel = Kernel(
data_reuploading_feature_map(num_features=4, reps=1, entanglement='full'),
'data_reuploading_feature_map(num_features=4, reps=1, entanglement=\'full\')',
)
params = dict(C=1.7782794100389228, kernel=kernel)
elif DATASET == 'downtown':
kernel = Kernel(
data_reuploading_feature_map(num_features=4, reps=1, entanglement='full'),
'data_reuploading_feature_map(num_features=4, reps=1, entanglement=\'full\')',
)
params = dict(C=10, kernel=kernel)
elif DATASET == 'ptgrey':
kernel = Kernel(
data_reuploading_feature_map(num_features=4, reps=1, entanglement='linear'),
'data_reuploading_feature_map(num_features=4, reps=1, entanglement=\'linear\')',
)
params = dict(C=10, kernel=kernel)
params |= kernel_svm_common_params
kernel_svm = SVC(**params)
kernel_svm.fit(train_x_normalized, train_y)
with open(LOG_DIR / f'kernel_svm_{DATASET}_ts={n_train_samples}.pkl', 'wb') as f:
dill.dump(kernel_svm, f)
with open(LOG_DIR / f'kernel_svm_{DATASET}_ts={n_train_samples}.pkl', 'rb') as f:
kernel_svm = dill.load(f)
preds = kernel_svm.predict(valid_x_normalized)
cm = confusion_matrix(preds, valid_y)
mcc = matthews_corrcoef(cm)
f1 = f1_score(cm)
acc = accuracy(cm)
print(f'Quantum Kernel SVM:\n\tcm = {cm.tolist()}\n\t{mcc = :.3f}\n\t{f1 = :.3f}\n\t{acc = :.2%}')
del kernel_svm
###################################################################################################################
# QSVM w/ Quantum Kernels #########################################################################################
###################################################################################################################
kernel_qsvm_sampler = 'hybrid'
kernel_qsvm_hybrid_time_limit = 120
kernel_qsvm_common_params = dict(
sampler=kernel_qsvm_sampler,
hybrid_time_limit=kernel_qsvm_hybrid_time_limit,
num_reads=1_000,
normalize=True,
threshold=0,
threshold_strategy='absolute',
optimize_memory=True,
dwave_api_token=None,
fail_to_classical=False,
)
if DATASET == 'kits':
kernel = Kernel(
data_reuploading_feature_map(num_features=4, reps=1, entanglement='full'),
'data_reuploading_feature_map(num_features=4, reps=1, entanglement=\'full\')',
)
params = dict(B=2, P=1, K=3, zeta=0.8, kernel=kernel)
elif DATASET == 'downtown':
kernel = Kernel(
data_reuploading_feature_map(num_features=4, reps=1, entanglement='linear'),
'data_reuploading_feature_map(num_features=4, reps=1, entanglement=\'full\')',
)
params = dict(B=2, P=1, K=3, zeta=0.4, kernel=kernel)
elif DATASET == 'ptgrey':
kernel = Kernel(
data_reuploading_feature_map(num_features=4, reps=1, entanglement='full'),
'data_reuploading_feature_map(num_features=4, reps=1, entanglement=\'full\')',
)
params = dict(B=2, P=1, K=3, zeta=0.4, kernel=kernel)
params |= kernel_qsvm_common_params
kernel_qsvm = QSVM(**params)
kernel_qsvm.fit(train_x, train_y)
with open(
LOG_DIR
/ f'kernel_qsvm_{DATASET}_ts={n_train_samples}_sampler={kernel_qsvm_sampler}_htl={kernel_qsvm_hybrid_time_limit}.pkl',
'wb',
) as f:
dill.dump(kernel_qsvm, f)
with open(
LOG_DIR
/ f'kernel_qsvm_{DATASET}_ts={n_train_samples}_sampler={kernel_qsvm_sampler}_htl={kernel_qsvm_hybrid_time_limit}.pkl',
'rb',
) as f:
kernel_qsvm = dill.load(f)
preds = kernel_qsvm.predict(valid_x)
cm = confusion_matrix(preds, valid_y)
mcc = matthews_corrcoef(cm)
f1 = f1_score(cm)
acc = accuracy(cm)
print(f'Quantum Kernel QSVM:\n\tcm = {cm.tolist()}\n\t{mcc = :.3f}\n\t{f1 = :.3f}\n\t{acc = :.2%}')
del kernel_qsvm
if __name__ == '__main__':
DATASET = sys.argv[1] if len(sys.argv) > 1 else None
assert DATASET is None or DATASET in ('kits', 'downtown', 'ptgrey')
WORKING_DIR = Path(__file__).parent
DATA_DIR = WORKING_DIR / 'logs' / 'ensemble_logs'
LOG_DIR = WORKING_DIR / 'logs' / f'large_logs_{DATASET}'
run_large()