-
Notifications
You must be signed in to change notification settings - Fork 0
/
tf-mnist-2-2-five_layers_relu_lrdecay_dropout.py
154 lines (126 loc) · 5.93 KB
/
tf-mnist-2-2-five_layers_relu_lrdecay_dropout.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
# -*- coding: utf-8 -*-
"""
Spyder Editor
This is a temporary script file.
"""
# encoding: UTF-8
# Copyright 2016 Google.com
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import tensorflow as tf
import pandas as pd
import matplotlib.pyplot as plt
import math
from tensorflow.examples.tutorials.mnist import input_data as mnist_data
print("Tensorflow version " + tf.__version__)
tf.set_random_seed(0)
RUNS = 10001
# prepare status updates, as the whole excercise may take quite some time
displays = 40 # how many time do you want to see the status
display_trigger = int(RUNS/displays)
# neural network with 5 layers
#
# · · · · · · · · · · (input data, flattened pixels) X [batch, 784] # 784 = 28*28
# \x/x\x/x\x/x\x/x\x/ -- fully connected layer (sigmoid) W1 [784, 200] B1[200]
# · · · · · · · · · Y1 [batch, 200]
# \x/x\x/x\x/x\x/ -- fully connected layer (sigmoid) W2 [200, 100] B2[100]
# · · · · · · · Y2 [batch, 100]
# \x/x\x/x\x/ -- fully connected layer (sigmoid) W3 [100, 60] B3[60]
# · · · · · Y3 [batch, 60]
# \x/x\x/ -- fully connected layer (sigmoid) W4 [60, 30] B4[30]
# · · · Y4 [batch, 30]
# \x/ -- fully connected layer (softmax) W5 [30, 10] B5[10]
# · Y5 [batch, 10]
# Download images and labels into mnist.test (10K images+labels) and mnist.train (60K images+labels)
mnist = mnist_data.read_data_sets("data", one_hot=True, reshape=False, validation_size=0)
# input X: 28x28 grayscale images, the first dimension (None) will index the images in the mini-batch
X = tf.placeholder(tf.float32, [None, 28, 28, 1])
# correct answers will go here
Y_ = tf.placeholder(tf.float32, [None, 10])
# variable learning rate
lr = tf.placeholder(tf.float32)
# Probability of keeping a node during dropout = 1.0 at test time (no dropout) and 0.75 at training time
pkeep = tf.placeholder(tf.float32)
# five layers and their number of neurons (tha last layer has 10 softmax neurons)
L = 200
M = 100
N = 60
O = 30
# Weights initialised with small random values between -0.2 and +0.2
# When using RELUs, make sure biases are initialised with small *positive* values for example 0.1 = tf.ones([K])/10
W1 = tf.Variable(tf.truncated_normal([784, L], stddev=0.1)) # 784 = 28 * 28
B1 = tf.Variable(tf.ones([L])/10)
W2 = tf.Variable(tf.truncated_normal([L, M], stddev=0.1))
B2 = tf.Variable(tf.ones([M])/10)
W3 = tf.Variable(tf.truncated_normal([M, N], stddev=0.1))
B3 = tf.Variable(tf.ones([N])/10)
W4 = tf.Variable(tf.truncated_normal([N, O], stddev=0.1))
B4 = tf.Variable(tf.ones([O])/10)
W5 = tf.Variable(tf.truncated_normal([O, 10], stddev=0.1))
B5 = tf.Variable(tf.zeros([10]))
# The model
# flatten the images into a single line of pixels
# -1 in the shape definition means "the only possible dimension that will preserve the number of elements"
XX = tf.reshape(X, [-1, 28*28])
# the layers
Y1 = tf.nn.relu(tf.matmul(XX, W1) + B1)
Y1d = tf.nn.dropout(Y1, pkeep)
Y2 = tf.nn.relu(tf.matmul(Y1d, W2) + B2)
Y2d = tf.nn.dropout(Y2, pkeep)
Y3 = tf.nn.relu(tf.matmul(Y2d, W3) + B3)
Y3d = tf.nn.dropout(Y3, pkeep)
Y4 = tf.nn.relu(tf.matmul(Y3d, W4) + B4)
Y4d = tf.nn.dropout(Y4, pkeep)
Ylogits = tf.matmul(Y4d, W5) + B5
Y = tf.nn.softmax(Ylogits)
# cross-entropy loss function (= -sum(Y_i * log(Yi)) ), normalised for batches of 100 images
# TensorFlow provides the softmax_cross_entropy_with_logits function to avoid numerical stability
# problems with log(0) which is NaN
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits=Ylogits, labels=Y_)
cross_entropy = tf.reduce_mean(cross_entropy)*100
# accuracy of the trained model, between 0 (worst) and 1 (best)
correct_prediction = tf.equal(tf.argmax(Y, 1), tf.argmax(Y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
# training, learning rate = 0.003
# training step, learning rate = 0.003
learning_rate = 0.003
train_step = tf.train.AdamOptimizer(lr).minimize(cross_entropy)
# init
init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init)
df = pd.DataFrame(columns=['train_accuracy', 'test_accuracy','train_cross_entropy', 'test_cross_entropy'])
# learning rate decay, preparation
max_learning_rate = 0.003
min_learning_rate = 0.0001
decay_speed = 2000.0 # 0.003-0.0001-2000=>0.9826 done in 5000 iterations
for i in range(RUNS):
if (i % display_trigger) == 0:
print(" run #" + str(i) + " of " + str(RUNS) + " runs.")
learning_rate = min_learning_rate + (max_learning_rate - min_learning_rate) * math.exp(-i/decay_speed)
batch_X, batch_Y = mnist.train.next_batch(100)
train_data = {X:batch_X, Y_: batch_Y, lr: learning_rate, pkeep: 0.75}
sess.run(train_step, feed_dict=train_data)
#success?
a, c = sess.run([accuracy, cross_entropy], feed_dict=train_data)
# success on test data?
test_data = {X : mnist.test.images, Y_ : mnist.test.labels, pkeep: 1.0}
at, ct = sess.run([accuracy, cross_entropy], feed_dict=test_data)
df.loc[len(df)] = [a, at, c, ct]
# display results
# display results
df1 = df[['train_accuracy', 'test_accuracy']]
df2 = df[['train_cross_entropy', 'test_cross_entropy']]
df1.plot()
df2.plot()
plt.show()