您的位置:首页 > 理论基础 > 计算机网络

《炼数成金》.第七课 递归神经网络LSTM的讲解,以及LSTM网络的使用学习笔记

2018-03-03 22:20 399 查看
笔记:
       1.递归神经网络和BP神经网络的区别:
                不同点:递归神经网络L有反馈回路,可以记住上一次的输出,并作为下一次的输入之一,BP神经网络没有反馈回路。
                相同点:都有梯度消失的问题,之前输入的数据会随着时间的流逝,信号会不断的衰弱,对决策的影响越来越小。

           使用y=x的激活函数,则不会出现梯度消失问题,但是网络会一直往下传播,重要的信息也记住,不重要的也记住。正确的应该正确的记住,错误的忘记,于是诞生了LSTM网络。
        2.权值和偏置值只设定一对就可以,其他没有设定的,tensorflow会自动为我们设置。
       3.代码调试过程中的错误:<ipython-input-8-a4f3de4a35a8> in RNN(X, weights, biases)
38 inputs = tf.reshape(X,[-1,max_time,n_inputs])
39 #定义LSTM基本CELL
---> 40 lstm_cell = tf.contrib.rnn.core_rnn_cell.BasicLSTMCell(lstm_size)
41 #lstm_cell = rnn.BasicLSTMCell(lstm_size)
42 # final_state[0]是cell state

AttributeError: module 'tensorflow.contrib.rnn' has no attribute 'core_rnn_cell'

     解决:
           #1.0版本改了很多
            #原代码是这样的:
            lstm_cell = tf.contrib.rnn.core_rnn_cell.BasicLSTMCell(lstm_size)
            #应该改为:
            from tensorflow.contrib import rnn

            lstm_cell = rnn.BasicLSTMCell(lstm_size) 
       4.最终输出准确率为93%,可以继续优化。源码
# coding: utf-8

# In[1]:

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
from tensorflow.contrib import rnn

# In[2]:

#载入数据集
mnist = input_data.read_data_sets("D://MNIST_data",one_hot=True)

# 输入图片是28*28
n_inputs = 28 #输入一行,一行有28个数据
max_time = 28 #一共28行
lstm_size = 100 #隐层单元
n_classes = 10 # 10个分类
batch_size = 50 #每批次50个样本
n_batch = mnist.train.num_examples // batch_size #计算一共有多少个批次

#这里的none表示第一个维度可以是任意的长度
x = tf.placeholder(tf.float32,[None,784])
#正确的标签
y = tf.placeholder(tf.float32,[None,10])

#初始化权值
weights = tf.Variable(tf.truncated_normal([lstm_size, n_classes], stddev=0.1))
#初始化偏置值
biases = tf.Variable(tf.constant(0.1, shape=[n_classes]))

#定义RNN网络
def RNN(X,weights,biases):
# inputs=[batch_size, max_time, n_inputs]
inputs = tf.reshape(X,[-1,max_time,n_inputs])
#定义LSTM基本CELL
#lstm_cell = tf.contrib.rnn.core_rnn_cell.BasicLSTMCell(lstm_size)
lstm_cell = rnn.BasicLSTMCell(lstm_size)
# final_state[0]是cell state
# final_state[1]是hidden_state
outputs,final_state = tf.nn.dynamic_rnn(lstm_cell,inputs,dtype=tf.float32)
results = tf.nn.softmax(tf.matmul(final_state[1],weights) + biases)
return results

#计算RNN的返回结果
prediction= RNN(x, weights, biases)
#损失函数
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=prediction,labels=y))
#使用AdamOptimizer进行优化
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
#结果存放在一个布尔型列表中
correct_prediction = tf.equal(tf.argmax(y,1),tf.argmax(prediction,1))#argmax返回一维张量中最大的值所在的位置
#求准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))#把correct_prediction变为float32类型
#初始化
init = tf.global_variables_initializer()

with tf.Session() as sess:
sess.run(init)
for epoch in range(6):
for batch in range(n_batch):
batch_xs,batch_ys =  mnist.train.next_batch(batch_size)
sess.run(train_step,feed_dict={x:batch_xs,y:batch_ys})

acc = sess.run(accuracy,feed_dict={x:mnist.test.images,y:mnist.test.labels})
print ("Iter " + str(epoch) + ", Testing Accuracy= " + str(acc))

# In[ ]:
输出
Extracting D://MNIST_data\train-images-idx3-ubyte.gz
Extracting D://MNIST_data\train-labels-idx1-ubyte.gz
Extracting D://MNIST_data\t10k-images-idx3-ubyte.gz
Extracting D://MNIST_data\t10k-labels-idx1-ubyte.gz
WARNING:tensorflow:From <ipython-input-7-89efd89d18dd>:52: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:

Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.

See tf.nn.softmax_cross_entropy_with_logits_v2.

Iter 0, Testing Accuracy= 0.8148
Iter 1, Testing Accuracy= 0.8737
Iter 2, Testing Accuracy= 0.899
Iter 3, Testing Accuracy= 0.9196
Iter 4, Testing Accuracy= 0.9217
Iter 5, Testing Accuracy= 0.9274
内容来自用户分享和网络整理,不保证内容的准确性,如有侵权内容,可联系管理员处理 点击这里给我发消息
标签: 
相关文章推荐