【文献阅读】Augmenting Supervised Neural Networks with Unsupervised Objectives-ICML-2016
一、Abstract
从近期对unsupervised learning 的研究得到启发,在large-scale setting 上,本文把unsupervised learning 与supervised learning结合起来,提高了supervised learning的性能。主要是把autoencoder与CNN结合起来
二、Key words:
SAE;SWWAE; reconstruction;encoder;decoder;VGG-16;Alex-Net
三、 Motivation
- reconstruction loss 很有用,reconstruction loss可以看作一个regularizer(SWWAE文中提到).
- unsupervised learning会对model起一定的限定作用,即相当于一个regularizer,这个regularizer使得encoder阶段提取得到的特征具有可解释性
四、Main contributions
- 本文实验表明了,high-capacity neural networks(采用了known switches)的 intermediate activations 可以保存input的大量信息,除了部分
2.通过结合decoder pathway 的loss,提升了supervised learning model的分类正确率
3.做了几个 autoencoder模型的对比实验,发现: the pooling switches and the layer-wise reconstruction loss 非常重要!
五、Inspired by
- Zhao, J., Mathieu, M., Goroshin, R., and Lecun, Y. Stacked what-where auto-encoders. ArXiv:1506.02351, 2015.
- Simonyan, K. and Zisserman, A. Very deep convolutional networks for large-scale image recognition. In ICLR,2015.
- Krizhevsky, A., Sutskever, I., and Hinton, G. E. Imagenet classification with deep convolutional neural networks.In NIPS, 2012.
Rasmus, A., Valpola, H., Honkala, M., Berglund, M., and Raiko, T. Semi-supervised learning with ladder network.In NIPS, 2015. - Adaptive deconvolutional networks for mid and high level feature learning
- Zeiler, M. D., Krishnan, D., Taylor, G. W., and Fergus, R. Deconvolutional networks. CVPR, 2010.
- Zeiler, M., Taylor, G., and Fergus, R. Adaptive deconvolu-tional networks for mid and high level feature learning.In ICCV, 2011.
key word:SWWAE;VGG-16;Alex-Net;ladder-Net;Deconvolutional network
六、文献具体实验及结果
1.SAE-all模型的训练:
第一步,采用VGG-16(训练好的VGG-16)初始化encoder,采用gaussian初始化decoder
第二步,固定encoder部分,用layerwise的方法训练decoder
第三步,用数据整体的训练更新decoder和encoder的参数
SAE-first模型的训练同SAE-all
SAE-layerwise一般只是拿来初始化 SAE-first SAE-all
SWWAE-all 提升了 1.66 % and 1.18% for single-crop and convolution schemes.
(top-1)
七、 感悟
- 2006~2010年期间, unsupervised learning 盛行是以为当时有标签数据不够大,所以需要用unsupervised leanring 的方法来初始化网络,可以取得较好效果,而 类似imagenet这样的大量标签数据的出现, 用autoencoder来初始化网络的优势已经没有。从这里也可以知道,当数据量较小时,可以考虑用unsupervised learning 的方法来初始化网络,从而提升分类准确率
- reconstruction loss 可以看作 regularization , 即是对enconder的weights做了一些限制,限制其获得的activations要能recon出input,是的提取得到的特征具有可解释性
【文献阅读】Augmenting Supervised Neural Networks with Unsupervised Objectives-ICML-2016的更多相关文章
- 【文献阅读】Self-Normalizing Neural Networks
Self-Normalizing Neural Networks ,长达93页的附录足以成为吸睛的地方(给人感觉很厉害), 此paper提出了新的激活函数,称之为 SELUs ,其具有normaliz ...
- 论文阅读 Streaming Graph Neural Networks
3 Streaming Graph Neural Networks link:https://dl.acm.org/doi/10.1145/3397271.3401092 Abstract 本文提出了 ...
- [ufldl]Supervised Neural Networks
要实现的部分为:forward prop, softmax函数的cost function,每一层的gradient,以及penalty cost和gradient. forwad prop forw ...
- [C3] Andrew Ng - Neural Networks and Deep Learning
About this Course If you want to break into cutting-edge AI, this course will help you do so. Deep l ...
- [Converge] Training Neural Networks
CS231n Winter 2016: Lecture 5: Neural Networks Part 2 CS231n Winter 2016: Lecture 6: Neural Networks ...
- An Intuitive Explanation of Convolutional Neural Networks
https://ujjwalkarn.me/2016/08/11/intuitive-explanation-convnets/ An Intuitive Explanation of Convolu ...
- 论文笔记之:Learning Multi-Domain Convolutional Neural Networks for Visual Tracking
Learning Multi-Domain Convolutional Neural Networks for Visual Tracking CVPR 2016 本文提出了一种新的CNN 框架来处理 ...
- How to Use Convolutional Neural Networks for Time Series Classification
How to Use Convolutional Neural Networks for Time Series Classification 2019-10-08 12:09:35 This blo ...
- 《Graph Neural Networks: A Review of Methods and Applications》阅读笔记
本文是对文献 <Graph Neural Networks: A Review of Methods and Applications> 的内容总结,详细内容请参照原文. 引言 大量的学习 ...
随机推荐
- PE笔记之NT头PE文件头
typedef struct _IMAGE_FILE_HEADER { WORD Machine; //014C-IMAGE_FILE ...
- 关于URL编码 [转]
转自: http://www.ruanyifeng.com/blog/2010/02/url_encoding.html 作者: 阮一峰 日期: 2010年2月11日 一.问题的由来 URL就是网址, ...
- (2)WCF客户端调用
一.visual studion引用生成代理 引入服务端发布元数据的地址(并不是服务的地址) 用服务端是控制台程序 例子1 服务端的配置 <system.serviceModel> < ...
- 2018年东北农业大学春季校赛 E 阶乘后的0【数论】
链接:https://www.nowcoder.com/acm/contest/93/E来源:牛客网 时间限制:C/C++ 1秒,其他语言2秒 空间限制:C/C++ 262144K,其他语言52428 ...
- 某考试T2 frog
题目背景 无 题目描述 数轴上有 n 只青蛙,分别编号为 1 到 n.青蛙 i 的初始位置的坐标为 xi. 它们准备进行如下形式的移动:每轮包括 m 次跳跃,第 i 次跳跃由青蛙 ai(1 < ...
- apache url rewrite问题
apache RewriteEngine Your browser sent a request that this server could not understand http://www.ra ...
- Blocks编程要点
[老狼推荐]Blocks编程要点原文:Blocks Programming Topics链接:http://developer.apple.com/library/ios/#documentation ...
- POJ 2128:Highways
Highways Time Limit: 2000MS Memory Limit: 65536K Total Submissions: 2730 Accepted: 1008 Specia ...
- java编程思想第四版第9章
练习3: public class MainTest { public static void main(String args[]){ Bcycle b=new Bcycle(); b.print( ...
- Windows API 教程(七) hook 钩子监听
茵蒂克丝 如何创建一个窗口 手动创建窗口的流程 实际代码 安装钩子 (Install hook) 钩子简介 SetWindowsHookEx 函数 设置监听[键盘]消息 设置监听[鼠标]消息 如何创建 ...