头歌-Tensorflow--构建手写字识别模型

第1关：Tensorflow-构建手写字识别模型

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

os.environ["TF_CPP_MIN_LOG_LEVEL"]='3'

'''设置输入层节点个数为784
 设置输出个数为10
 '''
#********* Begin *********#
INPUT_NODE = 784
OUTPUT_NODE = 10
#********* End  **********#

# 配置神经网络的参数
'''设置隐藏层节点个数为500，每轮喂入数据个数为100'''

LAYER1_NODE = 500
 
# 一次训练数据的个数
BATCH_SIZE = 100

#********* Begin *********#
# 隐藏层结点数
LAYER1_NODE = 500
 
# 一次训练数据的个数
BATCH_SIZE = 100

#********* End  **********#

LEARNING_RATE_BASE = 0.8  # 基础的学习率
LEARNING_RATE_DECAY = 0.99  # 学习率的衰减率
REGULARIZATION_RATE = 0.0001  # 正则化项的系数
TRAINING_STEPS = 3000  # 训练轮

def get_regularizer_variable(weights, regularizer):
    if regularizer != None:
        tf.add_to_collection("losses", regularizer(weights))
def forward(input_tensor, regularizer):
   #隐藏层参数定义与输出[784,500]
    weights1 = tf.Variable(tf.truncated_normal([INPUT_NODE, LAYER1_NODE], stddev=0.1))
    biases1 = tf.Variable(tf.constant(0.1, shape=[LAYER1_NODE]))
    get_regularizer_variable(weights1, regularizer)
    #input_tensor[None,784],[784,500]
    layer1 = tf.nn.relu(tf.matmul(input_tensor, weights1) + biases1)
   # 输出层参数与输出
    weights2 = tf.Variable(tf.truncated_normal([LAYER1_NODE, OUTPUT_NODE], stddev=0.1))
    biases2 = tf.Variable(tf.constant(0.1, shape=[OUTPUT_NODE]))
    get_regularizer_variable(weights2, regularizer)
    ouput = tf.matmul(layer1, weights2) + biases2
    return ouput
if __name__=="__main__":#程序运行的起点
    #读取mnist数据集[0,1,0,0,0,0,0,0,0,0]
    #********* Begin *********#
    mnist = input_data.read_data_sets("data", one_hot=True)

    #********* End  **********#

    #定义输入层节点，None指的是每次向神经网络中喂入图像的张数
    x = tf.placeholder(tf.float32, [None, INPUT_NODE], name='x-input')
    #神经网络的输出，有10个节点
    y_ = tf.placeholder(tf.float32, [None, OUTPUT_NODE], name='y-input')
    regularizer = tf.contrib.layers.l2_regularizer(REGULARIZATION_RATE)
    y= forward(x,regularizer)
    
    #计算预测值y与真实值y_的交叉熵
    #********* Begin *********#
    cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=y, labels=tf.argmax(y_,1))
    #********* End  **********#
    cross_entropy_mean = tf.reduce_mean(cross_entropy)  # 计算batch所有交叉熵的均值
    loss = cross_entropy_mean + tf.add_n(tf.get_collection("losses"))

    global_step = tf.Variable(0, trainable=False)

    #用指数衰减函数设置学习率
    #********* Begin *********#
    learning_rate = tf.train.exponential_decay(LEARNING_RATE_BASE,global_step,mnist.train.num_examples/BATCH_SIZE,LEARNING_RATE_DECAY)

    #********* End  **********#

    train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss)

    with tf.Session() as sess:
        init_op = tf.global_variables_initializer()
        sess.run(init_op)
        for i in range(TRAINING_STEPS):
            xs, ys = mnist.train.next_batch(BATCH_SIZE)
            _, _, _ = sess.run([train_step, loss, global_step], feed_dict={x: xs, y_: ys})
            if (i % 1000 == 0):
                sess.run(loss, feed_dict={x: xs, y_: ys})

    print("模型正确运行")