Dropout 层的python实现（内含dropout函数及带dropout的前向反向传播函数）

最新推荐文章于 2023-12-13 21:22:09 发布

Jillian_sea

最新推荐文章于 2023-12-13 21:22:09 发布

阅读量5.2k

点赞数

分类专栏：深度学习

本文链接：https://blog.csdn.net/Jillian_sea/article/details/80790212

版权

本文介绍了如何在Python中实现Dropout层，包括 Dropout 函数的详细解释以及带有 Dropout 的前向和反向传播过程。

摘要由CSDN通过智能技术生成

参考：https://zhuanlan.zhihu.com/p/29592806

# -*- coding:utf-8 -*-
import numpy as np
def dropout(X, keep_prob = 0.5):
    """
    :param X: input
    :param keep_prob:
    :return:
    """

    D = np.random.rand(X.shape[0], X.shape[1]) # step1: initialize matrix d
    D = D < keep_prob # step2: convert entries of d
    X = X * D # step3: shut down some neuron
    X = X/keep_prob #step4: scale the value of neuron

    return X

'''
numpy.random.rand(do,d1, ...,dn)
create an array of the given shape and populate it with random samples from a uniform distribution over[0,1)
'''

# 带 dropout的前向后向传播
def forward_propogation_with_dropout(X, parameters, keep_prob = 0.5):
    """
    Implemente the forward propagation : LINERE->