Coursera Deep Learning 2 Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization - week3, Hyperparameter tuning, Batch Normalization and Programming Frameworks
Tuning process
下图中的需要tune的parameter的先后顺序, 红色>黄色>紫色,其他基本不会tune.
先讲到怎么选hyperparameter, 需要随机选取(sampling at random)
随机选取的过程中,可以采用从粗到细的方法逐步确定参数

有些参数可以按照线性随机选取, 比如 n[l]

但是有些参数就不适合线性的sampling at radom, 比如 learning rate α,这时可以用 log


Andrew 很幽默的讲到了两种选参数的实际场景 pandas vs caviar. pandas approach 一般用在你的算力不够时候,要持续几天的training.

Batch norm
我们知道对input layer 做 normalizing, 其实对每一层的输入都可以做normalizing, 这就是 batch norm. 做batch norm 时,有对 activation后的结果做norm 的,也有对activation 前的结果 z 做batch norm 的,这里讲的是后一种,对z 做norm.




为什么Batch Norm 起作用呢?
先看下下面图讲到的convariate shift,如果traing set 的distribution 变了,就应该重新train model. 同样,对NN的每一层也有类似的问题.

Andrew讲到batch norm 是为了尽量使得不同layer decouple,这样相互影响就要小一点,整个NN比较稳定.

Batch norm 还有regularization 的作用,但是这个算法主要不是做这个的. 不建议专门用它来做regularization.

对 test set 求 μ, σ2, 采用了不一样的方法,就是基于签名mini-batch set 求出来的μ, σ2 应用exponetially weighted average 求平均值. 它和logistic regression 一样,decision boudary 是线性的.

Softmax Regression
Softmax regression 就是 logistic regression 的generaliazation 版本, 它可以用在multi-class clarification 问题上。和logistic regression 一样,decision boudary 都是线性的. 如果要使得decison boudary 是非线性的就需要deep network.



Programing framework
TensorFlow by google, an example

Coursera Deep Learning 2 Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization - week3, Hyperparameter tuning, Batch Normalization and Programming Frameworks的更多相关文章
- [C2W3] Improving Deep Neural Networks : Hyperparameter tuning, Batch Normalization and Programming Frameworks
第三周:Hyperparameter tuning, Batch Normalization and Programming Frameworks 调试处理(Tuning process) 目前为止, ...
- Coursera Deep Learning 2 Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization - week1, Assignment(Initialization)
声明:所有内容来自coursera,作为个人学习笔记记录在这里. Initialization Welcome to the first assignment of "Improving D ...
- Coursera Deep Learning 2 Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization - week1, Assignment(Gradient Checking)
声明:所有内容来自coursera,作为个人学习笔记记录在这里. Gradient Checking Welcome to the final assignment for this week! In ...
- Coursera Deep Learning 2 Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization - week1, Assignment(Regularization)
声明:所有内容来自coursera,作为个人学习笔记记录在这里. Regularization Welcome to the second assignment of this week. Deep ...
- Coursera Deep Learning 2 Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization - week2, Assignment(Optimization Methods)
声明:所有内容来自coursera,作为个人学习笔记记录在这里. 请不要ctrl+c/ctrl+v作业. Optimization Methods Until now, you've always u ...
- 课程二(Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization),第三周(Hyperparameter tuning, Batch Normalization and Programming Frameworks) —— 2.Programming assignments
Tensorflow Welcome to the Tensorflow Tutorial! In this notebook you will learn all the basics of Ten ...
- 吴恩达《深度学习》-课后测验-第一门课 (Neural Networks and Deep Learning)-Week 3 - Shallow Neural Networks(第三周测验 - 浅层神 经网络)
Week 3 Quiz - Shallow Neural Networks(第三周测验 - 浅层神经网络) \1. Which of the following are true? (Check al ...
- [CS231n-CNN] Training Neural Networks Part 1 : activation functions, weight initialization, gradient flow, batch normalization | babysitting the learning process, hyperparameter optimization
课程主页:http://cs231n.stanford.edu/ Introduction to neural networks -Training Neural Network ________ ...
- Coursera, Deep Learning 2, Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization - week1, Course
Train/Dev/Test set Bias/Variance Regularization 有下面一些regularization的方法. L2 regularation drop out da ...
随机推荐
- Java 并发集合的实现原理
http://www.codeceo.com/article/the-implementation-principle-of-java-concurrent-collection.html 阿凡卢 ...
- uoj#80 二分图最大权匹配
题意:给定二分图,有边权,求最大边权匹配.边权非负. 解:KM算法求解最大权完备匹配. 完备匹配就是点数少的那一边每个点都有匹配. 为了让完备匹配与最大权匹配等价,我们添加若干条0边使之成为完全二分图 ...
- 利用selenium并使用gevent爬取动态网页数据
首先要下载相应的库 gevent协程库:pip install gevent selenium模拟浏览器访问库:pip install selenium selenium库相应驱动配置 https: ...
- ImageMagickObject.MagickImage.1 '80041771' - 安装vc2008运行库
今天重装了系统后,发现 ImageMagick 不能正常的使用了: ImageMagickObject.MagickImage.1 '80041771' convert: 455: unable t ...
- 关于字符编码,你所需要知道的(ASCII,Unicode,Utf-8,GB2312…)
字符编码的问题看似很小,经常被技术人员忽视,但是很容易导致一些莫名其妙的问题.这里总结了一下字符编码的一些普及性的知识,希望对大家有所帮助. 还是得从ASCII码说起 说到字符编码,不得不说ASCII ...
- Spring入门初体验
Spring其实就是一个容器,让我们更方便的拿到想要的对象. 1.编写一个service // userService.java public interface userService { publ ...
- 《老梁四大名著情商课》笔记-学学TA,你就是聚会的万人迷
<老梁四大名著情商课>笔记-学学TA,你就是聚会的万人迷 作者:尹正杰 版权声明:原创作品,谢绝转载!否则将追究法律责任. 现在社会学家有一个统计,说中国处在单身状态大概有2个亿.这些人中 ...
- MapReduce-WordCount
pom.xml <?xml version="1.0" encoding="UTF-8"?> <project xmlns="htt ...
- Hadoop记录-日常运维操作
1.Active NameNode hang死,未自动切换 #登录当前hang死 Active namenode主机,停止Namenode,触发自动切换.hadoop-daemon.sh stop n ...
- svn Edge访问规则配置
权限的配置方法: 首先在用户页面定义用户,然后在版本库页面,手工填写访问规则. 以下面的配置为例(实际配置中应删除“//”及其后的注释): [groups] //定义用户组 ...