tf.nn.dropout

摘自：https://www.jianshu.com/p/c9f66bc8f96c

def dropout(x, keep_prob, noise_shape=None, seed=None, name=None)

输入是：

• x，你自己的训练、测试数据等
• keep_prob，dropout概率
• ……，其它参数不咋用

输出是：

• A Tensor of the same shape of x

输出的非0元素是原来的 “1/keep_prob” 倍

 

程序举例：

import tensorflow as tf

dropout = tf.placeholder(tf.float32)
x = tf.Variable(tf.ones([10, 10]))
y = tf.nn.dropout(x, dropout)

init = tf.initialize_all_variables()
sess = tf.Session()
sess.run(init)

print sess.run(y, feed_dict = {dropout: 0.4})

运行的结果如下：

[[ 0.   0.   2.5  2.5  0.   0.   2.5  2.5  2.5  2.5]
 [ 0.   2.5  2.5  2.5  2.5  2.5  0.   2.5  0.   2.5]
 [ 2.5  0.   0.   2.5  0.   0.   2.5  0.   2.5  0. ]
 [ 0.   2.5  2.5  2.5  2.5  0.   0.   2.5  0.   2.5]
 [ 0.   0.   0.   0.   0.   0.   0.   0.   2.5  2.5]
 [ 2.5  2.5  2.5  0.   2.5  0.   0.   2.5  2.5  2.5]
 [ 0.   2.5  2.5  2.5  0.   2.5  2.5  0.   0.   0. ]
 [ 0.   2.5  0.   2.5  0.   0.   2.5  2.5  0.   0. ]
 [ 2.5  2.5  2.5  2.5  2.5  0.   0.   2.5  0.   0. ]
 [ 2.5  0.   0.   0.   0.   0.   2.5  2.5  0.   2.5]]

分析一下运行结果：

• 输入和输出的tensor的shape果然是一样的
• 不是0的元素都变成了原来的 “1/keep_prob” 倍

tensorflow中的dropout就是：使输入tensor中某些元素变为0，其它没变0的元素变为原来的1/keep_prob大小！

博客作者说，小型网络效果一般，慎用