-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtl_learning_tests_TwoModality_double_net.py
179 lines (139 loc) · 10.1 KB
/
tl_learning_tests_TwoModality_double_net.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
# https://askubuntu.com/questions/8653/how-to-keep-processes-running-after-ending-ssh-session
# import warnings
# warnings.filterwarnings('ignore', r'All-NaN (slice|axis) encountered')
# warnings.simplefilter(action='ignore', category=FutureWarning)
from utils.utils import create_directory
from utils.utils import get_splited_list_of_files_and_scaler_HT, get_ht_specific_scalers
from utils.utils import Data_Generator, summrize, Data_Generator_Combiner
from classifiers.transferLearningDoubleNetAnomalyDetectorGroup import TLClassifierDouble
from sys import platform
import numpy as np
import os
os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID";
os.environ["CUDA_VISIBLE_DEVICES"]="1";
import tensorflow as tf
import keras
if __name__ == '__main__':
root_dir = 'HT_results/'
dir2bms_folder_refs = ['HT_Data/AES_withTrojan_Set1/',
'HT_Data/AES_withTrojan_Set7/']
dir2bms_folder_targets = ['HT_Data/AES_withTrojan_Set1/',
'HT_Data/AES_withTrojan_Set7/']
random_state = 10
classifier_name = 'htnetNarrowed'
base_folder='/transfer_learning_PWEM/'
use_ht_specific_scaler = True
use_refrence_scaler_for_traget_data= True
trained_models_folder= 'HT_trained_and_validated_on_all_EXCEPT_folder_name_different_classes'
tl_results_folder ='tl_results_with_double_net'
number_of_training_for_scaler = 100
number_of_batches_used_for_anomaly_detector_training = 50
lambda_ = 1
batch_size = 60
nb_epochs = 200
number_of_samples_per_folder = 1000
#Changes in the tests
changes_in_the_tests=True
circular_shift = 0
added_noise_mu = 0
added_noise_sigma = 0
# all_dataset_names =['AES-T400', 'AES-T500', 'AES-T600', 'AES-T700', 'AES-T800',
# 'AES-T1000', 'AES-T1100', 'AES-T1300', 'AES-T1400',
# 'AES-T1600', 'AES-T1800', 'AES-T2000']
# all_dataset_names = ['AES-T1600','AES-T1300', 'AES-T500', 'AES-T600', 'AES-T2000', 'AES-T400',
# 'AES-T700', 'AES-T800','AES-T1000', 'AES-T1100', 'AES-T1400', 'AES-T1800']
all_dataset_names = ['AES-T1100', 'AES-T500','AES-T1000' ,'AES-T700', 'AES-T800']#,'AES-T1600','AES-T1300',
# 'AES-T600', 'AES-T1400', 'AES-T1800', 'AES-T400','AES-T2000']
dataset_name_to_be_tested = all_dataset_names.copy()
# dataset_name_to_be_tested = ['AES-T1300','AES-T600', 'AES-T2000', 'AES-T500', 'AES-T400']
# dataset_name_to_be_tested = ['AES-T700', 'AES-T800',
# 'AES-T1000', 'AES-T1400',
# 'AES-T1600', 'AES-T1800']
# dataset_name_to_be_tested = ['AES-T1600','AES-T1300','AES-T1400', 'AES-T600', 'AES-T2000', 'AES-T500']
# dataset_name = 'AES-T700'
for dataset_name in dataset_name_to_be_tested:
trained_model_dir = root_dir +base_folder+ classifier_name + '/'+trained_models_folder + '/' + dataset_name+ '/'
output_directory = root_dir + base_folder + classifier_name + '/' + tl_results_folder + '/' + dataset_name+ '/'
if os.path.isdir(output_directory):
print('The results folder exists: ' + output_directory)
continue
create_directory(output_directory)
ref_generators, target_generators, val_generators, val_not_triggered_generators = [], [], [], []
for dir2bms_folder_ref, dir2bms_folder_target in zip(dir2bms_folder_refs, dir2bms_folder_targets):
dataset_names = all_dataset_names.copy()
# get not_triggered data training data from current chip
dirs_to_files_train, dirs_to_files_test, dirs_to_files_test_not_triggered, y_train, y_test, y_test_not_triggered, \
scaler, input_shape = get_splited_list_of_files_and_scaler_HT(dir2bms_folder= dir2bms_folder_target,
name_bms=[dataset_name], number_of_training_for_scaler= number_of_training_for_scaler,
use_trigerd_data_for_scale_training=False, random_state=random_state,
get_not_triggered_training_only=True, get_triggered_training_only=False,
get_not_triggered_validation_as_well = True,
number_of_samples_per_folder=number_of_samples_per_folder)
# get triggered and not triggered data from others
dataset_names.remove(dataset_name)
dirs_to_files_train_others, dirs_to_files_test_others, y_train_others, y_test_others, y_categorical_train_others, \
y_categorical_test_others, scaler_others, input_shape_others = \
get_splited_list_of_files_and_scaler_HT(dir2bms_folder= dir2bms_folder_ref, name_bms=dataset_names, number_of_training_for_scaler= number_of_training_for_scaler,
use_trigerd_data_for_scale_training= True,random_state=random_state,
get_not_triggered_training_only= False, get_triggered_training_only= False,
get_categorical_labels_y = True,number_of_samples_per_folder=number_of_samples_per_folder)
if use_refrence_scaler_for_traget_data:
scaler =scaler_others
if use_ht_specific_scaler:
scaler_others = get_ht_specific_scalers(dir2bms_folder_ref, dataset_names,
number_of_training_for_scaler=number_of_training_for_scaler)
ref_generators.append(Data_Generator(dirs_to_files_train_others, y_categorical_train_others,
dir2bms_folder = dir2bms_folder_ref, batch_size=batch_size, scaler=scaler_others))
target_generators.append(Data_Generator(dirs_to_files_train, y_train, batch_size=batch_size, scaler=scaler))
val_generators.append(Data_Generator(dirs_to_files_test, y_test, batch_size=batch_size, scaler=scaler,
circular_shift=circular_shift,added_noise_mu=added_noise_mu,
added_noise_sigma=added_noise_sigma))
val_not_triggered_generators.append(Data_Generator(dirs_to_files_test_not_triggered, y_test_not_triggered, batch_size=batch_size, scaler=scaler))
ref_generator = Data_Generator_Combiner(ref_generators)
target_generator = Data_Generator_Combiner(target_generators)
val_generator = Data_Generator_Combiner(val_generators)
val_not_triggered_generator = Data_Generator_Combiner(val_not_triggered_generators)
# nb_classes = len(np.unique(np.concatenate((y_train, y_test), axis =0)))
# classifier = transferLearning.Classifier_tlEncoder(output_directory, input_shape, nb_classes, verbose=1)
nb_classes = len(y_categorical_train_others[0])
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
config.gpu_options.per_process_gpu_memory_fraction = 0.8
keras.backend.tensorflow_backend.set_session(tf.Session(config=config))
classifier = TLClassifierDouble(base_model_dir=trained_model_dir, output_directory=output_directory,
nb_classes=nb_classes, batch_size=batch_size, lambda_=lambda_, verbose=1)
readme_file = open(output_directory + "/training_readme.txt", "w")
readme_file.write("batch_size: {}, nb_epochs: {},\n"
"Training Datasets: {}\n".format(batch_size, nb_epochs, str(dataset_names)))
readme_file.write("model_t: (target model)\n")
classifier.model_t.summary(print_fn=lambda x: readme_file.write(x + '\n'))
readme_file.write("\nmodel_r: (reference model)\n")
classifier.model_r.summary(print_fn=lambda x: readme_file.write(x + '\n'))
readme_file.close()
# classifier.fit_generator(ref_generator, target_generator, val_generator=val_generator,
# val_not_triggered_generator=val_not_triggered_generator,
# number_of_batches_used_for_anomaly_detector_training=
# number_of_batches_used_for_anomaly_detector_training,
# nb_epochs=nb_epochs,
# save_results_after_each_epoch=True, save_log_per_batch =True, expected_loss=None,
# patience_for_expected_loss=10)
classifier.evaluate_trained_feature_extractor(target_generator, val_generator,
number_of_batches_used_for_anomaly_detector_training)
# classifier.fit_generator_base(target_generator=target_generator, val_generator=val_generator,
# number_of_batches_used_for_anomaly_detector_training=
# number_of_batches_used_for_anomaly_detector_training)
# print('Classifier accuracy on current data: ', classifier.model.evaluate_generator(val_generator, workers=10, use_multiprocessing=True))
# print(scaler.mean_,scaler.scale_,len(scaler.scale_))
# print(scaler_others.mean_,scaler_others.scale_)
keras.backend.clear_session()
output_directory = root_dir + base_folder+ classifier_name + '/' + tl_results_folder + '/'
# summrize(output_directory, input_res='df_best_model.csv', output_res='all_best_model.csv')
added_txt=''
if changes_in_the_tests:
if circular_shift>0:
added_txt='_c_'+str(circular_shift)
if added_noise_sigma>0 or added_noise_mu>0:
added_txt=added_txt+'_mu_'+str(added_noise_mu)+'_sigma_'+str(added_noise_sigma)
summrize(output_directory, input_res='df_metrics.csv', output_res='all_metrics'+added_txt+'.csv')
print('Without feature extraction')
summrize(output_directory, input_res='df_metrics_base.csv', output_res='all_metrics_base'+added_txt+'.csv')