Tenserflow学习（二）——MNIST数据集分类三层网络搭建+Dropout+tensorboard可视化

1、上代码

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

# 载入数据
"""one_hot参数把标签转化到0-1之间
"""
mnist = input_data.read_data_sets("MNIST_data", one_hot=True)
# 每个批次大小（每次放入训练图像数量）
batch_size = 100
# 批次数量
num_batch = mnist.train.num_examples // batch_size


# 参数概要
def variable_summaries(var):
    with tf.name_scope('summaries'):
        mean = tf.reduce_mean(var)
        # 均值
        tf.summary.scalar('mean', mean)
        with tf.name_scope('stddev'):
            stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean)))
        # 标准差
        tf.summary.scalar('stddev', stddev)
        tf.summary.scalar('max', tf.reduce_max(var))
        tf.summary.scalar('min', tf.reduce_min(var))
        # 直方图
        tf.summary.histogram('histogram', var)


with tf.name_scope('input'):
    # placeholder
    x = tf.placeholder(tf.float32, [None, 784], name='x-input')
    y = tf.placeholder(tf.float32, [None, 10], name='y-input')

with tf.name_scope('keep_prob_And_learning_rate'):
    keep_prob = tf.placeholder(tf.float32, name='keep_prob')
    lr = tf.Variable(0.001, dtype=tf.float32, name='learning_rate')
    tf.summary.scalar('learning_rate', lr)

with tf.name_scope('layer'):
    with tf.name_scope('net-one'):
        with tf.name_scope('wights-one'):
            w1 = tf.Variable(tf.truncated_normal([784, 1000], stddev=0.1), name='w1')
            variable_summaries(w1)
        with tf.name_scope('biases-one'):
            b1 = tf.Variable(tf.zeros([1000]) + 0.1, name='b1')
            variable_summaries(b1)
        with tf.name_scope('drop-one'):
            L1 = tf.nn.tanh(tf.matmul(x, w1) + b1)
            L1_drop = tf.nn.dropout(L1, keep_prob)
    with tf.name_scope('net-two'):
        with tf.name_scope('wights-two'):
            w2 = tf.Variable(tf.truncated_normal([1000, 500], stddev=0.1), name='w2')
            variable_summaries(w2)
        with tf.name_scope('biases-two'):
            b2 = tf.Variable(tf.zeros([500]) + 0.1, name='b2')
            variable_summaries(b2)
        with tf.name_scope('drop-two'):
            L2 = tf.nn.tanh(tf.matmul(L1_drop, w2) + b2)
            L2_drop = tf.nn.dropout(L2, keep_prob)
    with tf.name_scope('net-three'):
        with tf.name_scope('wights-three'):
            w3 = tf.Variable(tf.truncated_normal([500, 100], stddev=0.1), name='w3')
            variable_summaries(w3)
        with tf.name_scope('biases-three'):
            b3 = tf.Variable(tf.zeros([100]) + 0.1, name='b3')
            variable_summaries(b3)
        with tf.name_scope('drop-three'):
            L3 = tf.nn.tanh(tf.matmul(L2_drop, w3) + b3)
            L3_drop = tf.nn.dropout(L3, keep_prob)
    with tf.name_scope('net-four-output'):
        with tf.name_scope('wights-four-final'):
            w4 = tf.Variable(tf.truncated_normal([100, 10], stddev=0.1), name='w4')
            variable_summaries(w4)
        with tf.name_scope('biases-four-final'):
            b4 = tf.Variable(tf.zeros([10]) + 0.1, name='b4')
            variable_summaries(b4)
        with tf.name_scope('prediction'):
            prediction = tf.nn.softmax(tf.matmul(L3_drop, w4) + b4)  # 概率值转化: softmax()

with tf.name_scope('loss'):
    # loss = tf.reduce_mean(tf.square(y - prediction))  # 二次代价函数
    loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(labels=y, logits=prediction))  # 交叉熵代价函数
    tf.summary.scalar('loss', loss)
with tf.name_scope('train'):
    # train_step = tf.train.GradientDescentOptimizer(0.2).minimize(loss)
    # train_step = tf.train.AdadeltaOptimizer(learning_rate=0.2, rho=0.95).minimize(loss)
    train_step = tf.train.AdamOptimizer(lr).minimize(loss)

init = tf.global_variables_initializer()
with tf.name_scope("accuracy-total"):
    with tf.name_scope("correct_prediction"):
        correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(prediction, 1))  # argmax()返回一维张量中最大值所在的位置
    # 计算准确率
    """cast()将correct_prediction列表变量中的值转换成float32 --> true=1.0，false=0.0
    """
    with tf.name_scope("accuracy"):
        accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))  # cast()相当于类型转换函数
        tf.summary.scalar('accuracy', accuracy)

# 合并summary
merged = tf.summary.merge_all()
with tf.Session() as sess:
    sess.run(init)
    writer = tf.summary.FileWriter('logs/', sess.graph)
    for epoch in range(51):
        sess.run(tf.assign(lr, 0.001 * (0.95 ** epoch)))  # tf.assign()属于赋值操作
        for batch in range(num_batch):
            batch_xs, batch_ys = mnist.train.next_batch(batch_size)
            summary, _ = sess.run([merged, train_step], feed_dict={x: batch_xs, y: batch_ys, keep_prob: 0.9})
        writer.add_summary(summary, epoch)
        test_acc = sess.run(accuracy, feed_dict={x: mnist.test.images, y: mnist.test.labels, keep_prob: 1.0})
        train_acc = sess.run(accuracy, feed_dict={x: mnist.train.images, y: mnist.train.labels, keep_prob: 1.0})
        learning_rate = sess.run(lr)
        print('iter' + str(epoch) + ', testing accuracy:' + str(test_acc) + ', training accuracy:' + str(train_acc)
              + ', learning rate:' + str(learning_rate))

2、dropout原理认识

参考：https://blog.csdn.net/program_developer/article/details/80737724

3、tensorboard可视化

" writer = tf.summary.FileWriter(‘logs/’, sess.graph) "记录了logs存储的路径，在所设置的路径下可以找到这样的文件：

打开终端进入到该文件位置，输入tensorboard --logdir=.\命令（比如我的文件在D:\PycharmProject\StudyDemo\logs下）：

在chrome中输入最后一行的：http:// … …
得到可视化数据和网络结构