拼命三娘冲（20191123)_Gradient checking

最新推荐文章于 2022-02-21 20:11:40 发布

coucoumoi

最新推荐文章于 2022-02-21 20:11:40 发布

阅读量108

点赞数

分类专栏：每天一小点文章标签：机器学习

本文链接：https://blog.csdn.net/weixin_38149396/article/details/103218581

版权

每天一小点专栏收录该内容

18 篇文章 0 订阅

订阅专栏

（吴恩达课程学习笔记）

Gradient checking that eliminates almost all of these problems, is useful when implement complex optimization models, and it will help you make sure and get sort of confident that your implementation of the algorithm is 100% correct.

To deal with little bugs which cannot be inspected, (e.g. it looks like the loss function is decreasing, however, you might just wind up with a neural network that has a higher level of error that you would with a bug-free implementation and you might just not know that was this subtle bug that gives worse performance.)