Tensorflow实现Softmax Regression 手写识别MNIST

来自＜Ｔｅｎｓｏｒｆｌｏｗ实战＞一书
# -*- coding:utf-8 -*-

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("MNIST_data/",one_hot=True)

print(mnist.train.images.shape, mnist.train.labels.shape)
print(mnist.test.images.shape, mnist.test.labels.shape)
print(mnist.validation.images.shape, mnist.validation.labels.shape)

sess = tf.InteractiveSession()

x = tf.placeholder(tf.float32,[None, 784])  #定义输入数据空间

W = tf.Variable(tf.zeros([784,10]))
b = tf.Variable(tf.zeros([10]))

y = tf.nn.softmax(tf.matmul(x,W) + b)

y_= tf.placeholder(tf.float32,[None, 10])  #定义输入标签空间
cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_*tf.log(y),reduction_indices=[1])) #定义损失函数
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(100)
    train_step.run({x: batch_xs, y_:batch_ys})

correct_prediction = tf.equal(tf.argmax(y, 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}))