使用tensorflow实现BP神经网络的Mnist手写体识别

【python—tensorflow】神经网络实现Mnist手写体的识别

#先导入识别所需要的包

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

 

mnist = input_data.read_data_sets(r"path",one_hot=True)

batch_size=100
n_batch=mnist.train.num_examples//batch_size
def variable_summaries(var):
    with tf.name_scope('summaries'):
        mean=tf.reduce_mean(var)
        tf.summary.scalar('mean',mean)#平均值
        with tf.name_scope('stddev'):
            stddev=tf.sqrt(tf.reduce_mean(tf.square(var - mean)))
        tf.summary.scalar('stddev',stddev)#标准差
        tf.summary.scalar('max',tf.reduce_max(var))#最大值
        tf.summary.scalar('min',tf.reduce_min(var))#最小值
        tf.summary.histogram('%histogram',var)#直方图

#数据与标签的占位
with tf.name_scope('input'):
    x = tf.placeholder(tf.float32,shape = [None,784],name='x-input')
    y = tf.placeholder(tf.float32,shape=[None,10],name='y-input')
   
    #keep_prob=tf.placeholder(tf.float32)
    #lr=tf.Variable(0.001,dtype=tf.float32)
with tf.name_scope('layer1'):
    #创建一个简单的神经网络
    with tf.name_scope('wight'):
        W1 = tf.Variable(tf.truncated_normal([784,500],stddev=0.1),name='W1')
        variable_summaries(W1)
    with tf.name_scope('biases'):
        b1=tf.Variable(tf.zeros([500])+0.1,name='b1')
        variable_summaries(b1)
        L1=tf.nn.tanh(tf.matmul(x,W1)+b1)

#第二层
with tf.name_scope('layer2'):
    with tf.name_scope('wight2'):
        W2=tf.Variable(tf.truncated_normal([500,10],stddev=0.1),name='W2')
        variable_summaries(W2)
    with tf.name_scope('biases2'):
        b2=tf.Variable(tf.zeros([10])+0.1,name='b2')
        variable_summaries(b2)
    with tf.name_scope('wx'):
        y_predict = tf.nn.softmax(tf.matmul(L1,W2) + b2)
#求交叉熵得到残差
with tf.name_scope('loss'):
    loss = tf.reduce_mean(tf.square(y-y_predict))#(这个与下面acc那个可以相互交替)
    tf.summary.scalar('loss',loss)
#梯度下降法使得残差最小
#train_step = tf.train.GradientDescentOptimizer(0.2).minimize(loss)
with tf.name_scope('train'):
    train_step = tf.train.GradientDescentOptimizer(0.2).minimize(loss)
    #loss=tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y,logits=y_predict))
#梯度下降法使得残差最小
#train_step = tf.train.Adam(lr).minimize(loss)
#测试阶段，测试准确度计算
with tf.name_scope('accuracy'):
    with  tf.name_scope('correct_prediction'):
        correct_prediction = tf.equal(tf.argmax(y_predict,1),tf.argmax(y,1))
    with  tf.name_scope('accuracy'):
        accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))#多个批次的准确度均值
        tf.summary.scalar('accuracy',accuracy)
    
#合并所有的summary
merged=tf.summary.merge_all()
init = tf.global_variables_initializer()
with tf.Session() as sess:
    sess.run(init)
    writer=tf.summary.FileWriter(r"C:\Users\Administrator\Desktop\logs",sess.graph)
    
    for i in range(100):
        for batch in range(n_batch):
            batch_xs,batch_ys = mnist.train.next_batch(batch_size)
            summary,_=sess.run([merged,train_step],feed_dict={x:batch_xs,y:batch_ys})
            #若想同时打印多个，则用中括号括起来
        writer.add_summary(summary,i)   
        test_acc=sess.run(accuracy,feed_dict={x: mnist.test.images, y: mnist.test.labels})
        train_acc=sess.run(accuracy,feed_dict={x: mnist.train.images, y: mnist.train.labels})
        print('Iter:'+ str(i)+ 'testing accuracy '+ str(test_acc) +'training accuracy '+ str(train_acc))