import numpy as np

import tensorflow as tf

import matplotlib.pyplot as plt

from tensorflow.examples.tutorials.mnist import input_data
# 导入数据

mnist = input_data.read_data_sets('data/', one_hot=True)

print (" tpye of 'mnist' is %s" % (type(mnist)))

print (" number of trian data is %d" % (mnist.train.num_examples))

print (" number of test data is %d" % (mnist.test.num_examples))

training = mnist.train.images

traininglable = mnist.train.labels

testing = mnist.test.images

testinglabel = mnist.test.labels

#初始化x和y

x = tf.placeholder('float', [None, 784])

y = tf.placeholder('float', [None, 10])

# 初始化W和b

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

b = tf.Variable(tf.zeros([10]))

sess = tf.Session()

# 构造多分类方程

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

# 构造代价函数y*log(y1), y1表示的是预测值

cost = tf.reduce_mean(-tf.reduce_sum(y*tf.log(actv), reduction_indices=1))

#训练模型

learning_rate = 0.01

#优化模型，使得cost最小化

optm = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)

# 预测结果的最大值索引与真实值的索引进行比对, tf.argmax( , 1) #找出一行中的最大值的索引

pred = tf.equal(tf.argmax(actv, 1), tf.argmax(y, 1))

# 计算正确率, tf.cast 把布尔值转换为数字形式

accr = tf.reduce_mean(tf.cast(pred, 'float'))

init = tf.global_variables_initializer()

# 训练次数

train_epoches = 50

# 每次抽取样本数

batch_size = 100

# 每5次循环打印一次结果

display_step = 5

sess = tf.Session()

sess.run(init)

for train_epoch in range(train_epoches):

    avg_cost = 0

    # 每次选取100个数据，循环的次数

    num_batch = int(mnist.train.num_examples/batch_size)

    for i in range(num_batch):

        # 抽取数据

        bacth_x, bacth_y = mnist.train.next_batch(batch_size)

        # 进行cost优化

        sess.run(optm, feed_dict={x:bacth_x, y:bacth_y})

        # 加上cost的值

        feeds = {x:bacth_x, y:bacth_y}

        avg_cost += sess.run(cost, feed_dict=feeds)/num_batch

    # 每5次打印一次结果

    if train_epoch % display_step == 0:

        feeds_train = {x:bacth_x, y:bacth_y}

        feed_test = {x:mnist.test.images, y:mnist.test.labels}

        # 计算训练集的准确率, feed_dict的参数

        train_acc = sess.run(accr, feed_dict=feeds_train)

        # 计算测试集的准确率

        test_acc = sess.run(accr, feed_dict=feed_test)

        print("Epoch: %03d/%03d cost: %.9f train_acc: %.3f test_acc: %.3f"

              % (train_epoch, train_epoches, avg_cost, train_acc, test_acc))