123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051
# -*- coding: utf-8 -*-"""Created on Sat Nov 25 09:03:11 2017@author: Administrator"""import input_datamnist = input_data.read_data_sets('MNIST',one_hot=True)import tensorflow as tf# 1/ placeholder表示一个占位符,将来运行可以传入新值x = tf.placeholder("float",[None,784]) # 2/ Variable表示一个可修改的张量,可以计算输入值也可以修改,一般用与模型参数# 3/ zeros[m,n] 可以生成 m,n 型矩阵,如 [2,3],int32) => [[0,0,0],[0,0,0],[0,0,0]]W = tf.Variable(tf.zeros([784,10]))b = tf.Variable(tf.zeros([10]))# 4/ nn为 NeuralNetwork 神经网络 , matmul 矩阵相乘y = tf.nn.softmax(tf.matmul(x,W)+b)y_ = tf.placeholder("float",[None,10])# 5/ reduce_sum() 计算一个张量的各个维度的元素的和。cross_entropy = -tf.reduce_sum(y_*tf.log(y))# 6/ GradientDescentOptimizer(0.01)梯度下降算法,0.01的学习速率最小化交叉熵train_step = tf.train.GradientDescentOptimizer(0.01).minimize(cross_entropy)init = tf.global_variables_initializer()sess = tf.Session()sess.run(init)# 训练模型for i in range(1000): batch_xs,batch_ys = mnist.train.next_batch(100) sess.run(train_step,feed_dict={x:batch_xs ,y_:batch_ys}) # 评估模型# 7/ tf.argmax(y,1)返回的是模型对于任一输入x预测到的标签值# 8/ tf.argmax(y_,1) 代表正确的标签correct_prediction = tf.equal(tf.argmax(y,1), tf.argmax(y_,1))# 9/ tf.reduce_mean()计算一个张量的各个维度的元素的平均值。# 10/ tf.cast()将一个张量投射到一个新的类型。accuracy = tf.reduce_mean(tf.cast(correct_prediction,"float"))print(sess.run(accuracy, feed_dict={x:mnist.test.images,y_:mnist.test.labels}))# 最终输出结果 0.9131