-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmodel_train.py
129 lines (92 loc) · 3.69 KB
/
model_train.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
import numpy as np
import numpy.matlib
import pandas as pd
import os
from sklearn.model_selection import train_test_split
from keras.models import Sequential
from keras.layers import Dense,Conv2D,Flatten,MaxPool2D,Dropout
from keras.callbacks import EarlyStopping, ModelCheckpoint
########################### set parameters #############################
path = '../path/to/data'
Kx = 11
Ky = 5
num_pca = 16
mode = 1 # 1 for CNN / 0 for MLP
######################### read data from file ##########################
def read_data(path):
Xs = []
ys = []
files = os.listdir(path)
for name in files:
data = pd.read_csv(os.path.join(path,name),header=None)
XX = data.iloc[:41,:].values
yy = data.iloc[41:,:].values
Xs.extend(XX.T)
ys.extend(yy.T)
return np.transpose(Xs) np.transpose(y)
X, y = read_data(path)
######################### create dataset for training ####################
def create_dataset(Xs,ys,Kx,Ky,mode):
kx = int((Kx-1)/2)
ky = int((Ky-1)/2)
# padd data to begining and end
XX = np.hstack((np.repeat(Xs[:,[0]],kx,axis=1), Xs, np.repeat(Xs[:,[-1]],kx,axis=1)))
yy = np.hstack((np.repeat(ys[:,[0]],ky,axis=1), ys, np.repeat(ys[:,[-1]],ky,axis=1)))
Xphn, ypca = ([] for i in range(2))
for i in range(kx,np.shape(XX)[1]-kx):
if mode:
tmp = XX[:,i-kx:i+kx+1]
tmp = tmp[:,np.newaxis,:]
Xphn.append(tmp)
else:
tmp = np.reshape(XX[:,i-kx:i+kx+1],(1,41*Kx),order='F')
Xphn.extend(tmp)
for j in range(ky,np.shape(yy)[1]-ky):
tmp = np.reshape(yy[:,j-ky:j+ky+1],(1,num_pca*Ky),order='F')
ypca.extend(tmp)
return np.array(Xphn), np.array(ypca)
Xphn, ypca = create_dataset(X,y,Kx,Ky,mode)
print(np.shape(Xphn))
print(np.shape(ypca))
def train_model(mode):
model = Sequential()
if mode:
model.add(Conv2D(256,kernel_size=(7,1),padding='same',activation='tanh',input_shape=(41,1,Kx)))
model.add(MaxPool2D(pool_size=(4,1),strides=(2,1)))
#model.add(Dropout(0.25))
model.add(Conv2D(512,kernel_size=(5,1),padding='same',activation='tanh'))
model.add(MaxPool2D(pool_size=(2,1),strides=(2,1)))
#model.add(Dropout(0.25))
#model.add(Conv2D(512,kernel_size=(3,1),padding='valid',activation='tanh'))
#model.add(MaxPool2D(pool_size=(2,1),strides=(2,1)))
#model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(3000,activation='tanh'))
model.add(Dropout(0.5))
model.add(Dense(3000,activation='tanh'))
model.add(Dropout(0.5))
model.add(Dense(16*Ky))
model.compile(loss='mse', optimizer='adam', metrics=['mse'])
else:
model = Sequential()
model.add(Dense(3000,activation='tanh',input_dim=41*Kx))
model.add(Dropout(0.5))
model.add(Dense(3000,activation='tanh'))
model.add(Dropout(0.5))
model.add(Dense(3000, activation='tanh'))
model.add(Dropout(0.5))
model.add(Dense(16*Ky))
model.compile(loss='mse',optimizer='adam',metrics=['mse'])
return model
######################## model train #######################################
Xtrn, Xtst, ytrn, ytst = train_test_split(Xphn,ypca,test_size=0.1,shuffle=True,random_state=42)
early_stop = EarlyStopping(monitor='val_loss', min_delta=0, patience=2, verbose=2, mode='auto')
mdl_save = ModelCheckpoint('mdl_wts.hdf5', save_best_only=True, monitor='val_loss', mode='auto')
model = train_model(mode)
model.fit(Xtrn,ytrn,validation_data=(Xtst,ytst),batch_size=128,epochs=10,shuffle=True,verbose=2,
callbacks=[early_stop, mdl_save])
### save model
model_json = model1.to_json()
with open("model.json", "w") as json_file:
json_file.write(model_json)
print("Model saved to disk")