-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy path07_lstm.py
121 lines (102 loc) · 5.1 KB
/
07_lstm.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
#-*- coding:utf-8 -*-
#!/usr/bin/env python
#Inspired by https://github.com/aymericdamien/TensorFlow-Examples/blob/master/examples/3%20-%20Neural%20Networks/recurrent_network.py
#LSTM 详解 http://blog.csdn.net/u014595019/article/details/52605693
# http://baijiahao.baidu.com/s?id=1579839249811513976&wfr=spider&for=pc
import tensorflow as tf
import numpy as np
## 单一 LSTM网络
#import input_data
# configuration
# O * W + b -> 10 labels for each image, O[? 28], W[28 10], B[10]
# ^ (O: output 28 vec from 28 vec input)
# |
# +-+ +-+ +--+
# |1|->|2|-> ... |28| time_step_size = 28
# +-+ +-+ +--+
# ^ ^ ... ^
# | | |
# img1:[28] [28] ... [28]
# img2:[28] [28] ... [28]
# img3:[28] [28] ... [28]
# ...
# img128 or img256 (batch_size or test_size 256)
# each input size = input_vec_size=lstm_size=28
'''
inputs 引数接受形态为 [batch_size,input_size] 的张量列表。
列表的长度为将网络展开后的时间步数，即列表中每一个元素都分别对应网络展开的时间步。
比如在 MNIST 数据集中，我们有 28x28 像素的图像，
每一张都可以看成拥有 28 行 28 个像素的图像。
我们将网络按 28 个时间步展开，以使在每一个时间步中，可以输入一行 28 个像素（input_size），
从而经过 28 个时间步输入整张图像。给定图像的 batch_size 值，则每一个时间步将分别收到 batch_size 个图像。
'''
## 变量配置
# configuration variables
input_vec_size = lstm_size = 28
# lstm_size 一个 LSTM 单元格中的单元数（标准LSTM单元）
time_step_size = 28# lstm 记录的时间步长
batch_size = 128#每一次训练 batch大小
test_size = 256 #每一次测试 batch大小
## 初始化权重 #####
def init_weights(shape):
return tf.Variable(tf.random_normal(shape, stddev=0.01))
## 模型 ##
def model(X, W, B, lstm_size):
# X, input shape: (batch_size, time_step_size, input_vec_size) 128 * 28 * 28
XT = tf.transpose(X, [1, 0, 2]) # 调换 0和1维度 转置 permute time_step_size and batch_size
# XT shape: (time_step_size, batch_size, input_vec_size)
XR = tf.reshape(XT, [-1, lstm_size]) # each row has input for each lstm cell (lstm_size=input_vec_size)
# XR shape: (time_step_size * batch_size, input_vec_size)
# 分割每行 得到每个时刻的输入
X_split = tf.split(0, time_step_size, XR) # split them to time_step_size (28 arrays)
# Each array shape: (batch_size, input_vec_size)
# Make lstm with lstm_size (each input vector size)
# TensorFlow 中，基础的 LSTM 单元格
lstm = tf.nn.rnn_cell.BasicLSTMCell(lstm_size, forget_bias=1.0, state_is_tuple=True)
# Get lstm cell output, time_step_size (28) arrays with lstm_size output: (batch_size, lstm_size)
outputs, _states = tf.nn.rnn(lstm, X_split, dtype=tf.float32)
# Linear activation
# Get the last output
return tf.matmul(outputs[-1], W) + B, lstm.state_size # State size to initialize the stat
## 数据处理
from tensorflow.examples.tutorials.mnist import input_data# 读取mnist手写字体数据工具
mnist = input_data.read_data_sets("../minist/MNIST_data/", one_hot=True)
# 训练数据 训练数据标签 测试数据 测试数据标签
trX, trY, teX, teY = mnist.train.images, mnist.train.labels, mnist.test.images, mnist.test.labels
# reshape成 lstm网络需要的 图片格式
trX = trX.reshape(-1, 28, 28)
teX = teX.reshape(-1, 28, 28)
## 训练数据的 占位符
X = tf.placeholder("float", [None, 28, 28])
Y = tf.placeholder("float", [None, 10])
# get lstm_size and output 10 labels
## 初始化权重和偏置
W = init_weights([lstm_size, 10])
B = init_weights([10])
## 模型
py_x, state_size = model(X, W, B, lstm_size)
## 计算代价函数 softmax回归 后 对数 信息增益 均值
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(py_x, Y))
## 训练优化操作
train_op = tf.train.RMSPropOptimizer(0.001, 0.9).minimize(cost)
## 预测值
predict_op = tf.argmax(py_x, 1)
## 启动 图 回话 Launch the graph in a session
with tf.Session() as sess:
# tf.initialize_all_variables() no long valid from
# 2017-03-02 if using tensorflow >= 0.12
if int((tf.__version__).split('.')[1]) < 12 and int((tf.__version__).split('.')[0]) < 1:
init = tf.initialize_all_variables()
else:
init = tf.global_variables_initializer()
sess.run(init)
for i in range(100):
for start, end in zip(range(0, len(trX), batch_size), range(batch_size, len(trX)+1, batch_size)):
# start=(0 128 128+128 ...) end =(128 128+128 128+128+128 ...)
sess.run(train_op, feed_dict={X: trX[start:end], Y: trY[start:end]})
test_indices = np.arange(len(teX)) # Get A Test Batch
np.random.shuffle(test_indices)
test_indices = test_indices[0:test_size]#测试数据
print(i, np.mean(np.argmax(teY[test_indices], axis=1) ==
sess.run(predict_op, feed_dict={X: teX[test_indices],
Y: teY[test_indices]})))#计算每次训练的正确率