Tensorflow1.0 第三集：Dropout

Tensorflow1.0 第三集：Dropout

---

防止过拟合的方法：

1. 增加数据集
2. 正则化方法：$C=C_0+\frac{\lambda }{2n}{\sum }_w{w}^2$，减少了网络的复杂程度
3. Dropout：训练的时候只用部分的神经元进行训练，测试的时候再将所有神经元激活

还是之前的手写数字集，继续加上Dropout：

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

# 输入数据集
mnist = input_data.read_data_sets("MNIST_data", one_hot=True)

# 每个批次大小，即每次训练的时候用的样本大小
batch_size = 50
# 计算一共有多少个批次
n_batch = mnist.train.num_examples // batch_size

x = tf.placeholder(tf.float32, [None, 784]) #每张图像有784个像素块 28*28的
y = tf.placeholder(tf.float32, [None, 10])

keep_prob = tf.placeholder(tf.float32) #dropout

# 一个神经网络
W = tf.Variable(tf.random_normal([784, 2000]))
b = tf.Variable(tf.zeros([1, 2000]))
L = tf.nn.tanh(tf.matmul(x, W) + b)
L_drop = tf.nn.dropout(L, keep_prob)

W1 = tf.Variable(tf.random_normal([2000, 2000]))
b1 = tf.Variable(tf.zeros([1, 2000]))
L1 = tf.nn.tanh(tf.matmul(L_drop, W1) + b1)
L1_drop = tf.nn.dropout(L1, keep_prob)

W2 = tf.Variable(tf.random_normal([2000, 10]))
b2 = tf.Variable(tf.random_normal([1, 10]))
L2 = tf.nn.tanh(tf.matmul(L1_drop, W2) + b2)
L2_drop = tf.nn.dropout(L2, keep_prob)

prediction = tf.nn.softmax(L2_drop)

# loss = tf.reduce_mean(tf.square(prediction - y))
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=prediction))
train_step = tf.train.GradientDescentOptimizer(learning_rate=0.2).minimize(loss)

init = tf.global_variables_initializer()

# 求准确率
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(prediction, 1)) #布尔型列表
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) #tf.cast 把布尔型转换为浮点型

with tf.Session() as sess:
    sess.run(init)
    for epoch in range(30):
        for batch in range(n_batch):
            batch_xs, batch_ys = mnist.train.next_batch(batch_size)
            sess.run(train_step, feed_dict={x: batch_xs, y: batch_ys, keep_prob: 0.7})

        test_acc = sess.run(accuracy, feed_dict={x: mnist.test.images, y: mnist.test.labels, keep_prob: 1.0}) # 测试的时候要让所有神经元都工作
        train_acc = sess.run(accuracy, feed_dict={x:mnist.train.images, y:mnist.train.labels, keep_prob: 1.0})
        print("Iter: " + str(epoch) + ", Testing Accuracy: " + str(test_acc) + ", Training Accuracy: " + str(train_acc))