tensorflow逻辑回归

# -*- coding: utf-8 -*-
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

__author__ = 'BrownWong'


# 获得训练数据
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
# 定义训练相关设置
training_epochs = 40  # 训练次数（一次训练读取一次训练集所有样本）
learning_rate = 0.01
batch_size = 100  # 一次训练中每batch个样本就更新一下参数
# 定义输入输出
x = tf.placeholder("float", [None, 784])
y = tf.placeholder("float", [None, 10])
# 定义参数
W = tf.Variable(tf.zeros([784, 10]))
b = tf.Variable(tf.zeros([10]))
# 定义cost function
activation = tf.nn.softmax(tf.matmul(x, W) + b)
cross_entropy = y*tf.log(activation)
cost = tf.reduce_mean(-tf.reduce_sum(cross_entropy, reduction_indices=1))  # 累计误差函数
# 定义优化算法
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)
# 定义参数初始化
init = tf.initialize_all_variables()
with tf.Session() as sess:
    sess.run(init)
    for epoch in range(training_epochs):
        sum_cost = 0.
        total_batch = int(mnist.train.num_examples / batch_size)
        for i in range(total_batch):
            batch_xs, batch_ys = mnist.train.next_batch(batch_size)
            sess.run(optimizer, feed_dict={x: batch_xs, y: batch_ys})
            sum_cost += sess.run(cost, feed_dict={x: batch_xs, y: batch_ys})
        print('Epoch: %02d  cost: %.9f' % (epoch + 1, sum_cost/mnist.train.num_examples))
    print("Training phase finished")
    # 模型评估
    correct_prediction = tf.equal(tf.argmax(activation, 1), tf.argmax(y, 1))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
    print("MODEL accuracy: %f" % accuracy.eval({x: mnist.test.images, y: mnist.test.labels}))