Tensorflow系列三：深度神经网络和MNIST数字识别

Tensorflow系列三：MNIST数字识别样例程序

1. mnist_inference定义前向传播函数及神经网络(单隐层)参数

#-*-coding:utf-8-*-
import tensorflow as tf
#定义神经网络结构
INPUT_NODE = 784
OUTPUT_NODE = 10
LAYER_NODE = 500
#通过get_variable函数获取变量，且将正则化加入损失集合
def get_weight_variable(shape,regularizer):
    #tf.get_variable与tf.Variable不同
    weights = tf.get_variable("weights",shape,
                              initializer=tf.truncated.normal_initializer(stddev=0.1))
    if regularizer != None:
        tf.add_to_collection('losses',regularizer(weights))
    return weights
#前向传播
def inference(input_tensor,regularizer):
    with tf.variable_scope('layer1'):
        weights = get_weight_variable([INPUT_NODE,LAYER_NODE],regulaerizer)
        bias = tf.get_variable('bias',[LAYER_NODE],tf.constant_initializer(0.0))
        #ReLu激活函数，类似还有tf.sigmoid和tf.tanh
        layer1 = tf.nn.relu(tf.matmul(input_tensor,weights)+bias)
    with tf.variable_scope('layer2'):
        weights = get_weight_variable([LAYER_NODE,OUTPUT_NODE],regulaerizer)
        bias = tf.get_variable('bias',[OUTPUT_NODE],tf.constant_initializer(0.0))
        layer2 = tf.matmul(layer1,weights)+bias
    return layer2

2. mnist_train训练程序

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

#配置神经网络训练相关参数
BATCH_SIZE = 100
LEARNING_RATE_BASE = 0.8
LEARNING_RATE_DECAY = 0.99
REGULARAZTION_RATE = 0.0001
TRAINING_STEPS = 30000
MOVING_AVERAGE_DECAY = 0.99
#模型持久化保存路径和文件名
MODEL_SAVE_PATH = '/path/to.model/'
MODE_NAME = "model.ckpt"

def train(mnist):
    x = tf.placeholder(tf.float32,[None,mnist.inference.INPUT_NODE],name="x-input")
    y_ = tf.placeholder(tf.float32,[None,mnist.inference.OUTPUT_NODE],name="y-input")
    regularizer = tf.contrib.layers.l2_regularizer(REGULARAZTION_RATE)
    y = mnist_inference.inference(x,regularizer)

    global_step = tf.Variable(0,trainable=False)
    #滑动平均
    variable_average = tf.ExponentialMovingAverage(MOVING_AVERAGE_DECAY,golbal_step)
    variable_average_op = variable_average.apply(tf.trainable_variables())
    #交叉熵
    cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(y,tf.argmax(y_,1))
    cross_entropy_mean = tf.reduce_mean(cross_entropy)
    loss = cross_entropy_mean+tf.add_n(tf.get_collection('losses'))
    #指数衰减学习率
    learning_rate = tf.train.exponential_decay(LEARNING_RATE_BASE,global_step,
                                              mnist.train.num_examples/BATCH_SIZE,
                                              LEARNING_RATE_DECAY)
    train_step = tf.train.GradientDescentOptimizer(learning_rate)
                   .minimize(loss,global_step=global_step)
    #一次完成多个操作：train_op = tf.group(train_step,variable_average_op)与下面等价
    with tf.control_dependencies([train_step, variable_average_op]):
        train_op = tf.no_op(name='train')
    #持久化
    saver = tf.train.Saver()
    with tf.Session() as sess:
        tf.initialize_all_variables().run()
        for i in range(TRAINING_STEPS):
            xs,ys = mnist.train.next_batch(BATCH_SIZE)
            _,loss_value,step = sess.run([train_op,loss,global_step],
                                        feed_dict={x:xs,y_:ys})
            if i%1000 ==0:
                print("After %d training steps, loss on batch is %g."%(step,loss_value))
                saver.save(sess,os.path.join(MODEL_SAVE_PATH,MODEL_NAME),golbal_step=global_step)
def main(argv=None):
    mnist = input_data.read_data_sets('/tmp/data',one_hot=True)
    train(mnist)
if __name__ == '__main__':
    tf.app.run()

3. mnist_eval 测试程序

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

#每10秒加载一次最新的模型，并在测试数据上测试最新模型的正确率
EVAL_INTERVAL_SECS = 10

def evaluate(mnist):
    with tf.Graph.as_default() as g:
        x = tf.placeholder(tf.float32,[None,mnist.inference.INPUT_NODE],name="x-input")
        y_ = tf.placeholder(tf.float32,[None,mnist.inference.OUTPUT_NODE],name="y-input")
        validate_feed = {x:mnist.validation.images,y_:mnist.validation.labels}

        y=mnist_inference.inference(x,None)
        correct_prediction = tf.equal(tf.argmax(y,1),tf.argmax(y_,1))
        accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))

        varidable_averages = tf.train.ExponentionMovingAverage(mnist_train.MOVING_AVERAGE_DECAY)
        variables_to_restore = variable_averages.variables_to_restore()
        saver = tf.train.Saver(variables_to_restore)

        while True:
            with tf.Session() as sess:
                #checkpoint文件维护了持久化的所有模型文件的文件名
                ckpt = tf.get_checkpoint_state(mnist_train.MODELSAVE_PATH)
                if ckpt and ckpt.model_checkpoint_path:
                    #model_checkpoint_path保存了最新的模型文件的文件名
                    saver.restore(sess,ckpt.model_checkpoint_path)
                    global_step = ckpt.model_checkpoint_path.split('/')[-].split('-')[-1]
                    accuracy_score = sess.run(accuracy,feed_dict=validate_feed)
                    print("After %d training steps, validation accuracy = %g."
                          %(step,accuracy_score))
                else:
                    print('No checkpoint file found')
                    return 
        time.sleep(EVAL_INTERCAL_SECS)
def main(argv=None):
    mnist = input_data.read_data_sets('/tmp/data',one_hot=True)
    evaluate(mnist)
if __name__ == '__main__':
    tf.app.run()