/
autoencoder.py
358 lines (290 loc) · 13.6 KB
/
autoencoder.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
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
# -*- coding: utf-8 -*-
""" Auto Encoder Example.
Using an auto encoder on MNIST handwritten digits.
References:
Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner. "Gradient-based
learning applied to document recognition." Proceedings of the IEEE,
86(11):2278-2324, November 1998.
Links:
[MNIST Dataset] http://yann.lecun.com/exdb/mnist/
================================How to run this script=================================
1. you can run the following command using DropNeuron
$ python autoencoder.py 0.00001 0 0.00001 1 0.005
You may have the following Summary of statistics:
$ sparsity of w_encoder_h1= 16.0036670918 %
$ sparsity of w_encoder_h2= 44.4702148438 %
$ sparsity of w_decoder_h1= 54.1137695312 %
$ sparsity of w_decoder_h2= 18.141143176 %
$ Total Sparsity= 42341 / 217088 = 19.5040720814 %
$ Compression Rate = 5.12713445596
$ NMSE without prune: 0.0121042
$ NMSE with prune: 0.0121184
$ Neuron percentage = 420 / 784 = 53.5714285714 %
$ Neuron percentage = 127 / 128 = 99.21875 %
$ Neuron percentage = 61 / 64 = 95.3125 %
$ Neuron percentage = 121 / 128 = 94.53125 %
$ Neuron percentage = 629 / 784 = 80.2295918367 %
$ Total Neuron Percentage = 1358 / 1888 = 71.9279661017 %
2. you can run the following command without Regularisation or DropOut
$ python autoencoder.py 0.00001 0 0 0.5 0.005
You may have the following Summary of statistics:
$ sparsity of w_encoder_h1= 15.1845503827 %
$ sparsity of w_encoder_h2= 46.2890625 %
$ sparsity of w_decoder_h1= 52.5268554688 %
$ sparsity of w_decoder_h2= 17.5392617985 %
$ Total Sparsity= 40934 / 217088 = 18.8559478184 %
$ Compression Rate = 5.30336639468
$ NMSE without prune: 0.0115187
$ NMSE with prune: 0.0115109
$ Neuron percentage = 459 / 784 = 58.5459183673 %
$ Neuron percentage = 127 / 128 = 99.21875 %
$ Neuron percentage = 62 / 64 = 96.875 %
$ Neuron percentage = 121 / 128 = 94.53125 %
$ Neuron percentage = 784 / 784 = 100.0 %
$ Total Neuron Percentage = 1553 / 1888 = 82.2563559322 %
$ python autoencoder.py 0 0 0 0.5 0.005
You may have the following Summary of statistics:
$ sparsity of w_encoder_h1= 99.5744977679 %
$ sparsity of w_encoder_h2= 99.3896484375 %
$ sparsity of w_decoder_h1= 99.4506835938 %
$ sparsity of w_decoder_h2= 99.6083785077 %
$ Total Sparsity= 216173 / 217088 = 99.5785119399 %
$ Compression Rate = 1.00423272102
$ NMSE without prune: 0.0311838
$ NMSE with prune: 0.0311848
$ Neuron percentage = 784 / 784 = 100.0 %
$ Neuron percentage = 128 / 128 = 100.0 %
$ Neuron percentage = 64 / 64 = 100.0 %
$ Neuron percentage = 128 / 128 = 100.0 %
$ Neuron percentage = 784 / 784 = 100.0 %
$ Total Neuron Percentage = 1888 / 1888 = 100.0 %
Author: Wei Pan
Contact: w.pan11@imperial.ac.uk
dropneuron@gmail.com
"""
import tensorflow as tf
from regularizers import *
import input_data
mnist = input_data.read_data_sets("/tmp/data/", one_hot=True)
import numpy as np
import sys
from scipy.io import savemat
import matplotlib.pyplot as plt
plt.ion()
lambda_l1 = float(sys.argv[1])
lambda_l2 = float(sys.argv[2])
lambda_dropneuron = float(sys.argv[3])
keep_prob = float(sys.argv[4]) # keep_prob \in (0, 1]
threshold = float(sys.argv[5])
learning_rate_ini = 0.001
training_epochs = 50
batch_size = 256
display_step = 1
examples_to_show = 10
# Network Parameters
n_hidden_1 = 128 # 1st layer num features
n_hidden_2 = 64 # 2nd layer num features
n_input = 784 # MNIST data input (img shape: 28*28)
# tf Graph input (only pictures)
input = tf.placeholder("float", [None, n_input])
W = {
'encoder_h1': tf.Variable(tf.random_normal([n_input, n_hidden_1], stddev=1)),
'encoder_h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2], stddev=1)),
'decoder_h1': tf.Variable(tf.random_normal([n_hidden_2, n_hidden_1], stddev=1)),
'decoder_h2': tf.Variable(tf.random_normal([n_hidden_1, n_input], stddev=1)),
}
W_prune = W
biases = {
'encoder_b1': tf.Variable(tf.random_normal([n_hidden_1], stddev=1)),
'encoder_b2': tf.Variable(tf.random_normal([n_hidden_2], stddev=1)),
'decoder_b1': tf.Variable(tf.random_normal([n_hidden_1], stddev=1)),
'decoder_b2': tf.Variable(tf.random_normal([n_input], stddev=1)),
}
# Building the encoder
def encoder(x):
layer_1 = tf.nn.sigmoid(tf.add(tf.matmul(x, W['encoder_h1']), biases['encoder_b1']))
tf.nn.dropout(layer_1, keep_prob)
layer_2 = tf.nn.sigmoid(tf.add(tf.matmul(layer_1, W['encoder_h2']), biases['encoder_b2']))
tf.nn.dropout(layer_2, keep_prob)
return layer_2
# Building the decoder
def decoder(x):
layer_1 = tf.nn.sigmoid(tf.add(tf.matmul(x, W['decoder_h1']), biases['decoder_b1']))
tf.nn.dropout(layer_1, keep_prob)
layer_2 = tf.nn.sigmoid(tf.add(tf.matmul(layer_1, W['decoder_h2']), biases['decoder_b2']))
tf.nn.dropout(layer_2, keep_prob)
return layer_2
def l1(x):
regularizers = (l1_regularizer(.1)(W['encoder_h1']) + l1_regularizer(.1)(biases['encoder_b1']))
regularizers += (l1_regularizer(.1)(W['encoder_h2']) + l1_regularizer(.1)(biases['encoder_b2']))
regularizers += (l1_regularizer(.1)(W['decoder_h1']) + l1_regularizer(.1)(biases['decoder_b1']))
regularizers += (l1_regularizer(.1)(W['decoder_h2']) + l1_regularizer(.1)(biases['decoder_b2']))
regularizers = x * regularizers
return regularizers
def l2(x):
regularizers = (l2_regularizer(.1)(W['encoder_h1']) + l2_regularizer(.1)(biases['encoder_b1']))
regularizers += (l2_regularizer(.1)(W['encoder_h2']) + l2_regularizer(.1)(biases['encoder_b2']))
regularizers += (l2_regularizer(.1)(W['decoder_h1']) + l2_regularizer(.1)(biases['decoder_b1']))
regularizers += (l2_regularizer(.1)(W['decoder_h2']) + l2_regularizer(.1)(biases['decoder_b2']))
regularizers = x * regularizers
return regularizers
def dropneuron(x):
regularizers = (lo_regularizer(.1)(W['encoder_h1'])) + tf.reduce_mean(li_regularizer(.1)(W['encoder_h1']))
regularizers += (lo_regularizer(.1)(W['encoder_h2'])) + tf.reduce_mean(li_regularizer(.1)(W['encoder_h2']))
regularizers += (lo_regularizer(.1)(W['decoder_h1'])) + tf.reduce_mean(li_regularizer(.1)(W['decoder_h1']))
regularizers += (lo_regularizer(.1)(W['decoder_h2'])) + tf.reduce_mean(li_regularizer(.1)(W['decoder_h2']))
regularizers = x * regularizers
return regularizers
def prune(x):
# Due to machine precision, typically, there is no absolute zeros solution.
# Therefore, we set a very small threshold to prune some parameters:
# However, the test error is obtained after pruning
y_noprune = sess.run(x)
y_noprune = np.asarray(y_noprune)
low_values_indices = abs(y_noprune) < threshold
y_prune = y_noprune
y_prune[low_values_indices] = 0
return y_noprune, y_prune
def neuron_input(w):
neuron_left = np.count_nonzero(np.linalg.norm(w, axis=1))
neuron_total = np.shape(w)[0]
print "Neuron percentage = ", neuron_left, "/", neuron_total, \
"=", float(neuron_left)/float(neuron_total)*100, "%"
return neuron_left, neuron_total
def neuron_output(w):
neuron_left = np.count_nonzero(np.linalg.norm(w, axis=0))
neuron_total = np.shape(w)[1]
print "Neuron percentage = ", neuron_left, "/", neuron_total, \
"=", float(neuron_left)/float(neuron_total)*100, "%"
return neuron_left, neuron_total
def neuron_layer(w1, w2):
neuron_in = np.count_nonzero(np.linalg.norm(w1, axis=0))
neuron_out = np.count_nonzero(np.linalg.norm(w2, axis=1))
neuron_left = min(neuron_in, neuron_out)
neuron_total = np.shape(w1)[1]
print "Neuron percentage = ", neuron_left, "/", neuron_total, \
"=", float(neuron_left)/float(neuron_total)*100, "%"
return neuron_left, neuron_total
# Construct model
encoder_op = encoder(input)
decoder_op = decoder(encoder_op)
y_pred = decoder_op
y_true = input
# Define loss and optimizer, minimize the squared error
loss = tf.reduce_mean(tf.pow(y_true - y_pred, 2))
cost = loss
cost += l1(lambda_l1)
cost += l2(lambda_l2)
cost += dropneuron(lambda_dropneuron)
optimizer = tf.train.AdamOptimizer(learning_rate_ini, beta1=0.9, beta2=0.999, epsilon=1e-08, use_locking=False).minimize(cost)
# Initializing the variables
init = tf.initialize_all_variables()
# Launch the graph
with tf.Session() as sess:
sess.run(init)
avg_cost = 0.
total_batch = int(mnist.train.num_examples/batch_size)
# Training cycle
for epoch in range(training_epochs):
# Loop over all batches
for i in range(total_batch):
batch_xs, batch_ys = mnist.train.next_batch(batch_size)
# Fit training using batch data
_, cost_value = sess.run([optimizer, cost], feed_dict={input: batch_xs})
loss_value = sess.run(loss, feed_dict={input: batch_xs})
# Display logs per epoch step
if epoch % display_step == 0:
print "Epoch:", '%04d' % (epoch+1), "loss=", "{:.9f}".format(loss_value)
print("Optimization Finished!")
nmse_noprune = sess.run(loss, feed_dict={input: mnist.test.images})
w_encoder_h1_, w_encoder_h1 = prune(W['encoder_h1'])
W_prune['encoder_h1'] = W['encoder_h1'].assign(w_encoder_h1, use_locking=False)
print "w_encoder_h1 =", '\n', w_encoder_h1, "shape = ", np.shape(w_encoder_h1)
w_encoder_h2_, w_encoder_h2 = prune(W['encoder_h2'])
W_prune['encoder_h2'] = W['encoder_h2'].assign(w_encoder_h2, use_locking=False)
print "w_encoder_h2 =", '\n', w_encoder_h2, "shape = ", np.shape(w_encoder_h2)
w_decoder_h1_, w_decoder_h1 = prune(W['decoder_h1'])
W_prune['decoder_h1'] = W['decoder_h1'].assign(w_decoder_h1, use_locking=False)
print "w_decoder_h1 =", '\n', w_decoder_h1, "shape = ", np.shape(w_decoder_h1)
w_decoder_h2_, w_decoder_h2 = prune(W['decoder_h2'])
W_prune['decoder_h2'] = W['decoder_h2'].assign(w_decoder_h2, use_locking=False)
print "w_decoder_h2 =", '\n', w_decoder_h2, "shape = ", np.shape(w_decoder_h2)
sess.run(W_prune)
sparsity = np.count_nonzero(w_encoder_h1)
sparsity += np.count_nonzero(w_encoder_h2)
sparsity += np.count_nonzero(w_decoder_h1)
sparsity += np.count_nonzero(w_decoder_h2)
print "sparsity of w_encoder_h1=", \
float(np.count_nonzero(w_encoder_h1))/float(np.size(w_encoder_h1))*100, "%"
print "sparsity of w_encoder_h2=", \
float(np.count_nonzero(w_encoder_h2))/float(np.size(w_encoder_h2))*100, "%"
print "sparsity of w_decoder_h1=", \
float(np.count_nonzero(w_decoder_h1))/float(np.size(w_decoder_h1))*100, "%"
print "sparsity of w_decoder_h2=", \
float(np.count_nonzero(w_decoder_h2))/float(np.size(w_decoder_h2))*100, "%"
num_parameter = np.size(w_encoder_h1)
num_parameter += np.size(w_encoder_h2)
num_parameter += np.size(w_decoder_h1)
num_parameter += np.size(w_decoder_h2)
total_sparsity = float(sparsity)/float(num_parameter)
print "Total Sparsity= ", sparsity, "/", num_parameter, \
" = ", total_sparsity*100, "%"
print "Compression Rate = ", float(num_parameter)/float(sparsity)
nmse_prune = sess.run(loss, feed_dict={input: mnist.test.images})
print "NMSE without prune:", nmse_noprune
print "NMSE with prune:", nmse_prune
# Applying encode and decode over test set
encode_decode = sess.run(y_pred, feed_dict={input: mnist.test.images[:examples_to_show]})
neuron_left_ = 0
neuron_total_ = 0
neuron_left, neuron_total = neuron_input(w_encoder_h1)
neuron_left_ += neuron_left
neuron_total_ += neuron_total
neuron_left, neuron_total = neuron_layer(w_encoder_h1, w_encoder_h2)
neuron_left_ += neuron_left
neuron_total_ += neuron_total
neuron_left, neuron_total = neuron_layer(w_encoder_h2, w_decoder_h1)
neuron_left_ += neuron_left
neuron_total_ += neuron_total
neuron_left, neuron_total = neuron_layer(w_decoder_h1, w_decoder_h2)
neuron_left_ += neuron_left
neuron_total_ += neuron_total
neuron_left, neuron_total = neuron_output(w_decoder_h2)
neuron_left_ += neuron_left
neuron_total_ += neuron_total
print "Total Neuron Percentage = ", \
neuron_left_, "/", neuron_total_, "=", float(neuron_left_)/float(neuron_total_)*100, "%"
savemat('result/result_autoencoder_dropout_only.mat',
{'w_encoder_h1_': w_encoder_h1_,
'w_encoder_h2_': w_encoder_h2_,
'w_decoder_h1_': w_decoder_h1_,
'w_decoder_h2_': w_decoder_h2_,
'w_encoder_h1': w_encoder_h1,
'w_encoder_h2': w_encoder_h2,
'w_decoder_h1': w_decoder_h1,
'w_decoder_h2': w_decoder_h2,
'learning_rate': learning_rate_ini,
'lambda_l1': lambda_l1,
'lambda_l2': lambda_l2,
'lambda_dropneuron': lambda_dropneuron,
'keep_prob': keep_prob,
'threshold': threshold,
'accuracy_prune': nmse_prune,
'accuracy_noprune': nmse_noprune})
# Compare original images with their reconstructions
f1, a1 = plt.subplots(1, 10, figsize=(10, 1))
for i in range(examples_to_show):
# a1[i].imshow(np.reshape(mnist.test.images[i], (28, 28)), cmap='Greys_r')
a1[i].imshow(np.reshape(mnist.test.images[i], (28, 28)))
a1[i].axis('off')
f1.show()
plt.savefig('m1.pdf', format='pdf')
f2, a2 = plt.subplots(1, 10, figsize=(10, 1))
for i in range(examples_to_show):
# a2[i].imshow(np.reshape(encode_decode[i], (28, 28)), cmap='Greys_r')
a2[i].imshow(np.reshape(encode_decode[i], (28, 28)))
a2[i].axis('off')
f2.show()
plt.savefig('m2.5.pdf', format='pdf')
plt.draw()
plt.waitforbuttonpress()