-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy path41c_amp_knets.py
318 lines (232 loc) · 11.3 KB
/
41c_amp_knets.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
import os
os.environ["CUDA_VISIBLE_DEVICES"]="0"
seed = 0
os.environ['PYTHONHASHSEED'] = str(seed)
import pickle
import numpy as np
import tensorflow as tf
import random as rn
np.random.seed(seed)
rn.seed(seed)
tf.random.set_seed(seed)
# #put the these lines before importing any module from keras.
gpus = tf.config.experimental.list_physical_devices('GPU')
for gpu in gpus:
tf.config.experimental.set_memory_growth(gpu, True)
from keras.callbacks import Callback
from keras.callbacks import ReduceLROnPlateau
from keras.callbacks import EarlyStopping
from keras import backend
from keras.optimizers import Adam, SGD
from sklearn.preprocessing import MinMaxScaler, StandardScaler
# from tensorflow.keras import initializers
from keras.models import Sequential
from keras.layers import Dense
import matplotlib.pyplot as plt
from time import time
start = time()
# monitor the learning rate
class LearningRateMonitor(Callback):
# start of training
def on_train_begin(self, logs={}):
self.lrates = list()
# end of each training epoch
def on_epoch_end(self, epoch, logs={}):
# get and store the learning rate
lrate = float(backend.get_value(self.model.optimizer.lr))
self.lrates.append(lrate)
with open('q1to8_s0.99_both/amp_rel_sur/tol_1e-10.pkl', 'rb') as f:
[lambda_values, coeffs, eim_basis, eim_indices] = pickle.load(f)
# SPLITTING TO K-NETWORKS
# K=1 for q [1, 4)
mask_train_k1 = (lambda_values[:,0] < 4.2)
lambda_values_k1 = lambda_values[mask_train_k1]
# K=2 for q [4,8]
mask_train_k2 = (lambda_values[:,0] >= 3.8)
lambda_values_k2 = lambda_values[mask_train_k2]
#TRAINING DATASET
# K=1 Network
lambda_values_k1[:, 0] = np.log10(lambda_values_k1[:, 0])
scaler_x_k1 = StandardScaler()
lambda_values_k1_scaled = scaler_x_k1.fit_transform(lambda_values_k1)
amp_train_x_k1 = lambda_values_k1_scaled
amp_train_y_k1 = coeffs[mask_train_k1,:]
ind_train_k1_first = np.where(lambda_values[:, 0] < 3.8)[0]
ind_train_overlap = np.where((lambda_values[:, 0] >= 3.8) & (lambda_values[:, 0] < 4.2))[0]
ind_train_k2_first = np.where(lambda_values[:, 0] >= 4.2)[0]
print(amp_train_x_k1.shape)
print(amp_train_y_k1.shape)
# K=2 Network
lambda_values_k2[:, 0] = np.log10(lambda_values_k2[:, 0])
scaler_x_k2 = StandardScaler()
lambda_values_k2_scaled = scaler_x_k2.fit_transform(lambda_values_k2)
amp_train_x_k2 = lambda_values_k2_scaled
amp_train_y_k2 = coeffs[mask_train_k2,:]
print(amp_train_x_k2.shape)
print(amp_train_y_k2.shape)
amp_train_x = np.zeros((200000,3))
amp_train_x[mask_train_k1,:] = amp_train_x_k1
amp_train_x[mask_train_k2,:] = amp_train_x_k2
# VALIDATION DATASET
with open('q1to8_s0.99_both/amp_rel_sur/tol_1e-10_val.pkl', 'rb') as f:
[lambda_values_val, coeffs_val, eim_basis_val, eim_indices_val] = pickle.load(f)
amp_val_y_all = coeffs_val
# K=1 for q [1, 4)
mask_val_k1 = (lambda_values_val[:,0] < 4.2)
lambda_values_val_k1 = lambda_values_val[mask_val_k1]
coeffs_val_k1 = coeffs_val[mask_val_k1, :]
ind_val_k1_first = np.where(lambda_values_val[:, 0] < 3.8)[0]
ind_val_overlap = np.where((lambda_values_val[:, 0] >= 3.8) & (lambda_values_val[:, 0] < 4.2))[0]
lambda_values_val_k1_log = lambda_values_val_k1.copy()
lambda_values_val_k1_log[:, 0] = np.log10(lambda_values_val_k1_log[:, 0])
lambda_values_val_k1_scaled = scaler_x_k1.transform(lambda_values_val_k1_log)
amp_val_x_k1 = lambda_values_val_k1_scaled
amp_val_y_k1 = coeffs_val[mask_val_k1,:]
print(amp_val_x_k1.shape)
print(amp_val_y_k1.shape)
# K=2 for q [4,8]
mask_val_k2 = (lambda_values_val[:,0] >= 3.8)
lambda_values_val_k2 = lambda_values_val[mask_val_k2]
ind_val_k2_first = np.where(lambda_values_val[:, 0] >= 4.2)[0]
lambda_values_val_k2_log = lambda_values_val_k2.copy()
lambda_values_val_k2_log[:, 0] = np.log10(lambda_values_val_k2_log[:, 0])
lambda_values_val_k2_scaled = scaler_x_k2.transform(lambda_values_val_k2_log)
amp_val_x_k2 = lambda_values_val_k2_scaled
amp_val_y_k2 = coeffs_val[mask_val_k2,:]
print(amp_val_x_k2.shape)
print(amp_val_y_k2.shape)
amp_val_x = np.zeros((30000,3))
amp_val_x[mask_val_k1,:] = amp_val_x_k1
amp_val_x[mask_val_k2,:] = amp_val_x_k2
# K=1 NETWORK
amp_model_k1 = Sequential()
amp_model_k1.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_model_k1.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_model_k1.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_model_k1.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_model_k1.add(Dense(18, activation=None))
# compile model
opt_k1 = Adam(lr=0.001)
amp_model_k1.compile(loss='mse', optimizer=opt_k1, metrics=['mse'])
amp_rlrp_k1 = ReduceLROnPlateau(monitor='mse', factor=0.9 , patience=15)
amp_lrm_k1 = LearningRateMonitor()
amp_stop_k1 = EarlyStopping(monitor='mse', patience=1000)
amp_history_k1 = amp_model_k1.fit(amp_train_x_k1, amp_train_y_k1, validation_data=(amp_val_x_k1, amp_val_y_k1),
epochs=1000, batch_size=256,
verbose=1, callbacks=[amp_rlrp_k1, amp_lrm_k1, amp_stop_k1])
#thelw kai predictions apo validaion
amp_val_predictions_k1 = amp_model_k1.predict(amp_val_x_k1)
amp_val_mse_k1 = np.square(amp_val_y_k1-amp_val_predictions_k1).mean()
print("Amplitude validation before residual mse K=1: ", amp_val_mse_k1)
# K=2 NETWORK
amp_model_k2 = Sequential()
amp_model_k2.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_model_k2.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_model_k2.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_model_k2.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_model_k2.add(Dense(18, activation=None))
# compile model
opt_k2 = Adam(lr=0.001)
amp_model_k2.compile(loss='mse', optimizer=opt_k2, metrics=['mse'])
amp_rlrp_k2 = ReduceLROnPlateau(monitor='mse', factor=0.9 , patience=15)
amp_lrm_k2 = LearningRateMonitor()
amp_stop_k2 = EarlyStopping(monitor='mse', patience=1000)
amp_history_k2 = amp_model_k2.fit(amp_train_x_k2, amp_train_y_k2, validation_data=(amp_val_x_k2, amp_val_y_k2),
epochs=1000, batch_size=256,
verbose=1, callbacks=[amp_rlrp_k2, amp_lrm_k2, amp_stop_k2])
amp_val_predictions_k2 = amp_model_k2.predict(amp_val_x_k2)
amp_val_mse_k2 = np.square(amp_val_y_k2 - amp_val_predictions_k2).mean()
print("Amplitude validation before residual mse K=2: ", amp_val_mse_k2)
amp_final_mse = (amp_val_mse_k1 + amp_val_mse_k2)/2
print("AMplitude final mse before residual: ", amp_final_mse)
lambda_train_overlap = lambda_values[ind_train_overlap]
lambda_train_overlap_log = lambda_train_overlap.copy()
lambda_train_overlap_log[: ,0] = np.log10(lambda_train_overlap_log[: ,0])
lambda_overlap_log_train_scaled_k1 = scaler_x_k1.transform(lambda_train_overlap_log)
lambda_overlap_log_train_scaled_k2 = scaler_x_k2.transform(lambda_train_overlap_log)
amp_train_predictions_k1 = amp_model_k1.predict(amp_train_x[ind_train_k1_first])
amp_train_predictions_o1 = amp_model_k1.predict(lambda_overlap_log_train_scaled_k1)
amp_train_predictions_k2 = amp_model_k2.predict(amp_train_x[ind_train_k2_first])
amp_train_predictions_o2 = amp_model_k2.predict(lambda_overlap_log_train_scaled_k2)
amp_train_predictions_overlap = (amp_train_predictions_o1 + amp_train_predictions_o2)/2
amp_train_final_predictions = np.zeros((200000,18))
amp_train_final_predictions[ind_train_k1_first,:] = amp_train_predictions_k1
amp_train_final_predictions[ind_train_overlap,:] = amp_train_predictions_overlap
amp_train_final_predictions[ind_train_k2_first,:] = amp_train_predictions_k2
#SAVE VALIDATION PREDICTIONS
lambda_overlap = lambda_values_val[ind_val_overlap]
lambda_overlap_log = lambda_overlap.copy()
lambda_overlap_log[: ,0] = np.log10(lambda_overlap_log[: ,0])
lambda_overlap_log_scaled_k1 = scaler_x_k1.transform(lambda_overlap_log)
lambda_overlap_log_scaled_k2 = scaler_x_k2.transform(lambda_overlap_log)
amp_val_predictions_k1 = amp_model_k1.predict(amp_val_x[ind_val_k1_first])
amp_val_predictions_o1 = amp_model_k1.predict(lambda_overlap_log_scaled_k1)
amp_val_predictions_k2 = amp_model_k2.predict(amp_val_x[ind_val_k2_first])
amp_val_predictions_o2 = amp_model_k2.predict(lambda_overlap_log_scaled_k2)
amp_val_predictions_overlap = (amp_val_predictions_o1 + amp_val_predictions_o2)/2
amp_val_final_predictions = np.zeros((30000,18))
amp_val_final_predictions[ind_val_k1_first,:] = amp_val_predictions_k1
amp_val_final_predictions[ind_val_overlap,:] = amp_val_predictions_overlap
amp_val_final_predictions[ind_val_k2_first,:] = amp_val_predictions_k2
amp_train_final_predictions = np.float64(amp_train_final_predictions)
amp_errors = coeffs - amp_train_final_predictions
amp_val_final_predictions = np.float64(amp_val_final_predictions)
amp_errors_val = coeffs_val - amp_val_final_predictions
# input for the residual error network
amp_res_scaler = MinMaxScaler()
amp_errors_scaled = amp_res_scaler.fit_transform(amp_errors)
amp_errors_val_scaled = amp_res_scaler.transform(amp_errors_val)
# Residual Error
amp_res_model = Sequential()
amp_res_model.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_res_model.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_res_model.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_res_model.add(Dense(320, activation=tf.keras.layers.ReLU()))
amp_res_model.add(Dense(18, activation=None))
opt_res = Adam(lr=0.001)
amp_res_model.compile(loss='mse', optimizer=opt_res, metrics=['mse'])
amp_res_rlrp = ReduceLROnPlateau(monitor='mse', factor=0.9, patience=50)
amp_res_lrm = LearningRateMonitor()
amp_res_stop = EarlyStopping(monitor='mse', patience=50)
amp_res_history = amp_res_model.fit(amp_train_x, amp_errors_scaled, validation_data=(amp_val_x, amp_errors_val_scaled),
epochs=50, batch_size=256,
verbose=1,
callbacks=[amp_res_rlrp, amp_res_lrm, amp_res_stop])
# Plot training & validation accuracy values
plt.figure()
plt.plot(amp_res_history.history['mse'])
plt.plot(amp_res_history.history['val_mse'])
plt.yscale("log")
plt.title('Residual Amp Model MSE')
plt.ylabel('mse')
plt.xlabel('Epoch')
plt.legend(['Train', 'Validation'], loc='upper right')
plt.savefig('Fig3: Amp Res Model Train MSE')
plt.show()
# Learning Rate
# plt.figure()
# plt.plot(amp_res_lrm.lrates)
# plt.yscale("log")
# plt.title('Amp Residual Learning Rate')
# plt.xlabel('Epoch')
# plt.savefig('Fig4: Amp Res Model Learning Rate')
# plt.show()
#
# # FINAL MSE
#
# predictions from residual and reverse transform to the INITIAL data space
amp_train_y_error = amp_res_scaler.inverse_transform(amp_res_model.predict(amp_train_x))
amp_val_y_error = amp_res_scaler.inverse_transform(amp_res_model.predict(amp_val_x))
amp_train_predictions_corrected = amp_train_final_predictions + amp_train_y_error
amp_val_final_predictions_with_res = amp_val_final_predictions + amp_val_y_error
amp_final_train_mse = np.square(amp_train_predictions_corrected - coeffs).mean()
amp_final_val_mse = np.square(amp_val_final_predictions_with_res - coeffs_val).mean()
print('Amplitude')
print("Amplitude validation before residual mse K=1: ", amp_val_mse_k1)
print("Amplitude validation before residual mse K=2: ", amp_val_mse_k2)
print("AMplitude final mse before residual: ", amp_final_mse)
print('Validation mse after residual: ', amp_final_val_mse)
# SAVE PREDICTIONS FOR MISMATCH SCRIPT
with open('Amplitude_nn_predictions#5.pkl', 'wb') as f:
pickle.dump([amp_val_final_predictions, amp_val_final_predictions_with_res], f)
print('minutes:', (time()-start)/60)