-
Notifications
You must be signed in to change notification settings - Fork 0
/
HyperModel.py
214 lines (169 loc) · 7.11 KB
/
HyperModel.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
from __future__ import print_function
import numpy as np
# random seeds must be set before importing keras & tensorflow
import Utils
my_seed = 123
np.random.seed(my_seed)
import random
random.seed(my_seed)
import tensorflow as tf
tf.random.set_seed(my_seed)
import csv
from hyperopt import Trials, STATUS_OK, tpe, hp, fmin
import hyperopt
from keras.layers import Input, Dense, Dropout
from keras.models import Model
from keras.utils import np_utils
from keras import callbacks
from sklearn.metrics import confusion_matrix
from keras.optimizers import RMSprop, Adadelta, Adagrad, Nadam, Adam
import global_config
from sklearn.model_selection import train_test_split
import time
import keras.backend as K
SavedParameters = []
def NN(x_train, y_train, params):
print(params)
input_shape = (x_train.shape[1],)
print(input_shape)
input = Input(input_shape)
l1 = Dense(params['neurons1'], activation='relu', kernel_initializer='glorot_uniform')(input)
l1 = Dropout(params['dropout1'])(l1)
# l1= BatchNormalization()(l1)
l1 = Dense(params['neurons2'], activation='relu', kernel_initializer='glorot_uniform')(
l1)
l1 = Dropout(params['dropout2'])(l1)
l1 = Dense(params['neurons3'], activation='relu', kernel_initializer='glorot_uniform')(
l1)
softmax = Dense(global_config.n_class, activation='softmax', kernel_initializer='glorot_uniform')(l1)
adam = Adam(lr=params['learning_rate'])
model = Model(inputs=input, outputs=softmax)
#model.summary()
model.compile(loss='categorical_crossentropy', metrics=['accuracy'],
optimizer=adam)
callbacks_list = [
callbacks.EarlyStopping(monitor='val_loss', min_delta=0.0001, patience=10,
restore_best_weights=True),
]
XTraining, XValidation, YTraining, YValidation = train_test_split(x_train, y_train, stratify=y_train,
test_size=0.2) # before model building
tic = time.time()
h = model.fit(x_train, y_train,
batch_size=params["batch"],
epochs=150,
verbose=2,
callbacks=callbacks_list,
validation_data=(XValidation, YValidation))
toc = time.time()
time_tot=toc-tic
scores = [h.history['val_loss'][epoch] for epoch in range(len(h.history['loss']))]
score = min(scores)
#print(score)
y_test = np.argmax(YValidation, axis=1)
Y_predicted = model.predict(XValidation, verbose=0, use_multiprocessing=True, workers=12)
Y_predicted = np.argmax(Y_predicted, axis=1)
cf = confusion_matrix(y_test, Y_predicted)
return model, h, {"val_loss": score , "time": time_tot,
"TP_val": cf[0][0],
"FN_val": cf[0][1], "FP_val": cf[1][0], "TN_val": cf[1][1]
}
def fit_and_score(params):
global SavedParameters
y_train = np_utils.to_categorical(global_config.train_Y, global_config.n_class)
model, h, val = NN(global_config.train_X, y_train, params)
#print(val)
print("start predict")
Y_predicted = model.predict(global_config.test_X, verbose=0, use_multiprocessing=True, workers=12)
Y_predicted = np.argmax(Y_predicted, axis=1)
elapsed_time = val['time']
cf = confusion_matrix(global_config.test_Y, Y_predicted)
#print(cf)
K.clear_session()
SavedParameters.append(val)
#print(SavedParameters)
global best_val_acc
SavedParameters[-1].update(
{
"learning_rate": params["learning_rate"],
"batch": params["batch"],
"dropout1": params["dropout1"],
"dropout2": params["dropout2"],
"neurons_layer1": params["neurons1"],
"neurons_layer2": params["neurons2"],
"neurons_layer3": params["neurons3"],
"time": time.strftime("%H:%M:%S", time.gmtime(elapsed_time))
})
cm_val = [[SavedParameters[-1]["TP_val"], SavedParameters[-1]["FN_val"]],
[SavedParameters[-1]["FP_val"], SavedParameters[-1]["TN_val"]]]
r = Utils.getResult(cm_val, global_config.n_class)
SavedParameters[-1].update({
"OA_val": r[4],
"P_val": r[6],
"R_val": r[7],
"F1_val": r[8],
"FAR_val": r[9],
"TPR_val": r[10]
})
SavedParameters[-1].update({
"TP_test": cf[0][0],"FN_test": cf[0][1], "FP_test": cf[1][0], "TN_test": cf[1][1]
})
cm_test = [[SavedParameters[-1]["TP_test"], SavedParameters[-1]["FN_test"]],
[SavedParameters[-1]["FP_test"], SavedParameters[-1]["TN_test"]]]
r = Utils.getResult(cm_test, False)
SavedParameters[-1].update({
"OA_test": r[4],
"P_test": r[6],
"R_test": r[7],
"F1_test": r[8],
"FAR_test": r[9],
"TPR_test": r[10]
})
# Save model
if SavedParameters[-1]["F1_val"] > global_config.best_accuracy:
print("new saved model:" + str(SavedParameters[-1]))
global_config.best_model=model
global_config.best_accuracy= SavedParameters[-1]["F1_val"]
best_score = SavedParameters[-1]["F1_val"]
print("Best score " + str(best_score))
'''
if SavedParameters[-1]["F1_test"] > best_test_acc:
print("new saved model Test:" + str(SavedParameters[-1]))
model.save(Name.replace(".csv", "_Test_model.h5"))
best_test_acc = SavedParameters[-1]["F1_test"]
'''
SavedParameters = sorted(SavedParameters, key=lambda i: i['F1_val'], reverse=True)
try:
with open(global_config.test_path + 'Results.csv', 'w', newline='') as csvfile:
writer = csv.DictWriter(csvfile, fieldnames=SavedParameters[0].keys())
writer.writeheader()
writer.writerows(SavedParameters)
except IOError:
print("I/O error")
return {'loss': -best_score, 'status': STATUS_OK} # cambia
def reset_global_variables(train_X, train_Y, test_X, test_Y):
global_config.train_X = train_X
global_config.train_Y = train_Y
global_config.test_X = test_X
global_config.test_Y = test_Y
global_config.best_score = 0
global_config.best_scoreTest = 0
global_config.best_accuracy=0
global_config.best_model = None
global_config.best_model_test = None
global_config.best_time = 0
def hypersearch(train_X, train_Y, test_X, test_Y, testPath, n_class):
reset_global_variables(train_X, train_Y, test_X, test_Y)
global_config.n_class=n_class
global_config.test_path = testPath
space = {"batch": hp.choice("batch", [32, 64, 128, 256, 512]),
'dropout1': hp.uniform("dropout1", 0, 1),
'dropout2': hp.uniform("dropout2", 0, 1),
"learning_rate": hp.loguniform("learning_rate", np.log(0.0001), np.log(0.01)),
"neurons1": hp.choice("neurons1", [128, 256, 512]),
"neurons2": hp.choice("neurons2", [64, 128, 256]),
"neurons3": hp.choice("neurons3", [32, 64, 128])}
trials = Trials()
best = fmin(fit_and_score, space, algo=tpe.suggest, max_evals=20, trials=trials,
rstate=np.random.RandomState(my_seed))
#best_params = hyperopt.space_eval(space, best)
return global_config.best_model, global_config.best_time