forked from zackchase/learningActivations
/
z_alpha_node_3c_2h.py
170 lines (121 loc) · 5.25 KB
/
z_alpha_node_3c_2h.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
import theano
from theano import tensor as T
from theano.sandbox.rng_mrg import MRG_RandomStreams as RandomStreams
import numpy as np
from load import mnist, cifar
from theano.tensor.nnet.conv import conv2d
from theano.tensor.signal.downsample import max_pool_2d
srng = RandomStreams()
np.random.seed()
def floatX(X):
return np.asarray(X, dtype=theano.config.floatX)
def init_weights(shape):
return theano.shared(floatX(np.random.randn(*shape) * 0.01))
def rectify(X):
return T.maximum(X, 0.)
def softmax(X):
e_x = T.exp(X - X.max(axis=1).dimshuffle(0, 'x'))
return e_x / e_x.sum(axis=1).dimshuffle(0, 'x')
def dropout(X, p=0.):
if p > 0:
retain_prob = 1 - p
X *= srng.binomial(X.shape, p=retain_prob, dtype=theano.config.floatX)
X /= retain_prob
return X
def reg_L2(w,w2,w3,w4,w5,w_o):
reg = T.sum(w**2)
reg += T.sum(w2**2)
reg += T.sum(w3**2)
reg += T.sum(w4**2)
reg += T.sum(w5**2)
reg += T.sum(w_o**2)
return 0.5*reg
def RMSprop(cost, params, lr=0.001, rho=0.9, epsilon=1e-6):
grads = T.grad(cost=cost, wrt=params)
updates = []
for p, g in zip(params, grads):
acc = theano.shared(p.get_value() * 0.)
acc_new = rho * acc + (1 - rho) * g ** 2
gradient_scaling = T.sqrt(acc_new + epsilon)
g = g / gradient_scaling
updates.append((acc, acc_new))
updates.append((p, p - lr * g ))
return updates
def model(X, w, w2, w3, w4, w5, w_o, b_h1, b_h2, b_o, p_drop_conv, p_drop_hidden):
l1_lin = conv2d(X, w, border_mode='full')+b_c1.dimshuffle('x', 0, 'x', 'x')
l1a = alpha_c1 * rectify(l1_lin) + (1.- alpha_c1) * T.tanh(l1_lin)
l1 = max_pool_2d(l1a, (2, 2))
l1 = dropout(l1, p_drop_conv)
l2_lin = conv2d(l1, w2) + b_c2.dimshuffle('x', 0, 'x', 'x')
l2a = alpha_c2 * rectify(l2_lin) + (1. - alpha_c2) * T.tanh(l2_lin)
l2 = max_pool_2d(l2a, (2, 2))
l2 = dropout(l2, p_drop_conv)
l3_lin = conv2d(l2, w3) + b_c3.dimshuffle('x', 0, 'x', 'x')
l3a = alpha_c3 * rectify(l3_lin) + ( 1 - alpha_c3) * T.tanh(l3_lin)
l3b = max_pool_2d(l3a, (2, 2))
l3 = T.flatten(l3b, outdim=2)
l3 = dropout(l3, p_drop_conv)
l4_lin = T.dot(l3, w4) + b_h1
l4 = alpha_h1 * rectify(l4_lin) + (1.-alpha_h1) * T.tanh(l4_lin)
l4 = dropout(l4, p_drop_hidden)
l5_lin = T.dot(l4, w5) + b_h2
l5 = alpha_h1 * rectify(l5_lin) + (1.-alpha_h2) * T.tanh(l5_lin)
l5 = dropout(l5, p_drop_hidden)
pyx = softmax(T.dot(l5, w_o) + b_o )
return l1, l2, l3, l4, l5, pyx
# load mnist data
# trX, teX, trY, teY = mnist(onehot=True)
# trX = trX.reshape(-1, 1, 28, 28)
# teX = teX.reshape(-1, 1, 28, 28)
# load cifar data
trX, teX, trY, teY = cifar(onehot=True)
trX = trX.reshape((-1, 3, 32, 32))
teX = teX.reshape((-1, 3, 32, 32))
print "trX shape: "
print trX.shape
X = T.ftensor4()
Y = T.fmatrix()
w = init_weights((32, 3, 3, 3))
w2 = init_weights((64, 32, 3, 3))
w3 = init_weights((128, 64, 3, 3))
w4 = init_weights((128 * 3 * 3, 625))
w5 = init_weights((625, 625))
w_o = init_weights((625, 10))
b_c1 = theano.shared(floatX(np.zeros(32)))
b_c2 = theano.shared(floatX(np.zeros(64)))
b_c3 = theano.shared(floatX(np.zeros(128)))
b_h1 = theano.shared(floatX(np.zeros(625)))
b_h2 = theano.shared(floatX(np.zeros(625)))
b_o = theano.shared(floatX(np.zeros(10)))
alpha_c1 = theano.shared(floatX(np.asarray([.5,])), broadcastable=[True] )
alpha_c2 = theano.shared(floatX(np.asarray([.5,])), broadcastable=[True] )
alpha_c3 = theano.shared(floatX(np.asarray([.5,])), broadcastable=[True] )
alpha_h1 = theano.shared(floatX(np.zeros(625) + .5))
alpha_h2= theano.shared(floatX(np.zeros(625) + .5))
noise_l1, noise_l2, noise_l3, noise_l4, noise_l5, noise_py_x = model(X, w, w2, w3, w4, w5, w_o, b_h1, b_h2, b_o, 0.5, 0.5)
l1, l2, l3, l4, l5, py_x = model(X, w, w2, w3, w4, w5, w_o, b_h1, b_h2, b_o, 0., 0.)
y_x = T.argmax(py_x, axis=1)
lambda_2 = .0005
cost = T.mean(T.nnet.categorical_crossentropy(noise_py_x, Y)) + lambda_2*reg_L2(w,w2,w3,w4,w5,w_o)
params = [w, w2, w3, w4, w5, w_o,
b_c1, b_c2, b_c3, b_h1, b_h2, b_o,
]
updates = RMSprop(cost, params, lr=0.001)
updates.append([alpha_c1, alpha_c1 - .001 * T.grad(cost, alpha_c1) ])
updates.append([alpha_c2, alpha_c2 - .001 * T.grad(cost, alpha_c2) ])
updates.append([alpha_c3, alpha_c3 - .001 * T.grad(cost, alpha_c3) ])
updates.append([alpha_h1, alpha_h1 - .001 * T.grad(cost, alpha_h1)])
updates.append([alpha_h2, alpha_h2 - .001 * T.grad(cost, alpha_h2)])
train = theano.function(inputs=[X, Y], outputs=cost, updates=updates, allow_input_downcast=True)
predict = theano.function(inputs=[X], outputs=y_x, allow_input_downcast=True)
def train_some_epochs(trX=trX, trY=trY, teX=teX, teY=teY, num_epochs=10):
for i in range(num_epochs):
####################################
# shuffle the rows before each epoch
####################################
p = np.random.permutation(len(trX))
trX = trX[p]
trY = trY[p]
for start, end in zip(range(0, len(trX), 128), range(128, len(trX), 128)):
cost = train(trX[start:end], trY[start:end])
print "test acc: " + str(np.mean(np.argmax(teY, axis=1) == predict(teX))) + "\ttrain acc:" + str(np.mean(np.argmax(trY[1:10000], axis=1) == predict(trX[1:10000])))