tensorflow 中tf.where，tf.greater梯度传播方式

在tensorflow中，tf.where 和tf.greater的组合相当于是一个分段函数，其梯度传播主要取决于里面不同condition，以下是一个简单示例：当x1小于0时，y1对x1的梯度为0，梯度在这里停止传播；当x2大于0时，y2对于x2的梯度为1；

 

# -*- coding: utf-8 -*-
"""
Created on Wed Dec  5 14:58:56 2018

@author: lunxi.yuan
"""

import tensorflow as tf 
with tf.Graph().as_default() as graph:
   


    x1 = tf.Variable(-1)
    x2 = tf.Variable(3)
   
    y1=tf.where(tf.greater(x1,0),x1,0)
    
    y2 = tf.where(tf.greater(x2,0),x2,0)

    grad_1 = tf.gradients(y1, x1)
    grad_2 = tf.gradients(y2, x2)
    with tf.Session(graph=graph) as sess:
        sess.run(tf.global_variables_initializer())
    
        print(sess.run(grad_1))
        print(sess.run(grad_2))