tensorflow 进阶(三),BP神经网络之两层hidden_layer

    本文与上一篇文章有一点不同,就是中间的隐藏层由一层变成两层,在神经网络搭建的过程中,曾出现一点问题,就是正确率图突然变成0.11,通过调整隐藏节点的数量和W2的初值,正确率达到0.97,不如只有一层神经网络的结果

1. 神经网络并不是越深越好
2. 神经网络的结构应当接近与橄榄型
3. 对于ML而言,不宜多于五层

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Thu Mar 22 22:15:25 2018

@author: luogan
"""
'''
from tensorflow.examples.tutorials.mnist import input_data

mnist=input_data.read_data_sets('MNIST_data/',one_hot=True)
'''

print(mnist.train.images.shape)

print(mnist.train.labels.shape)

print(mnist.test.images.shape)

print(mnist.test.images.shape)

a=mnist.train.images[8]
#a.reshape(28,28)
import pandas as pd
#b=pd.DataFrame(a)
#b
b=pd.DataFrame(a.reshape(28,28))
#b
#b=pd.DataFrame(a.reshape(28,28))
b.to_excel('c.xls')
d=mnist.train.labels[8]


print(mnist.validation.images.shape)
print(mnist.validation.labels.shape)

import tensorflow as tf
sess=tf.InteractiveSession()



in_units=784
h1_units=500

h2_units=100

w1=tf.Variable(  tf.truncated_normal([in_units,h1_units],stddev=0.1 ) )
b1=tf.Variable(tf.zeros([h1_units]))


w2=tf.Variable(tf.truncated_normal([h1_units,h2_units],stddev=0.1 ))
b2=tf.Variable(tf.zeros([h2_units]))

print('*'*20)

w3=tf.Variable(tf.zeros([h2_units,10]))
b3=tf.Variable(tf.zeros([10]))

x=tf.placeholder(tf.float32,[None,784])
keep_prob=tf.placeholder(tf.float32)


hidden1=tf.nn.relu(tf.matmul(x,w1)+b1)
hidden1_drop=tf.nn.dropout(hidden1,keep_prob)




hidden2=tf.nn.relu(tf.matmul(hidden1_drop,w2)+b2)

hidden2_drop=tf.nn.dropout(hidden2,keep_prob)

y2=tf.nn.softmax(tf.matmul(hidden2_drop,w3)+b3)



y_=tf.placeholder(tf.float32,[None,10])


cross_entropy=tf.reduce_mean(-tf.reduce_sum(y_*tf.log(y2),
                                            reduction_indices=[1]))

train_step=tf.train.GradientDescentOptimizer(0.5).minimize(cross_entropy)


tf.global_variables_initializer().run()

for i in range(1000):
    batch_xs,batch_ys=mnist.train.next_batch(1000)
    train_step.run({x:batch_xs,y_:batch_ys,keep_prob:1.0})



correct_prediction=tf.equal(tf.argmax(y2,1),tf.argmax(y_,1))

accuracy=tf.reduce_mean(tf.cast(correct_prediction,tf.float32))

print(accuracy.eval({x:mnist.test.images,y_:mnist.test.labels,keep_prob:1}))

(55000, 784)
(55000, 10)
(10000, 784)
(10000, 784)
(5000, 784)
(5000, 10)
********************
0.9775