何毓琦的个人博客分享 http://blog.sciencenet.cn/u/何毓琦 哈佛(1961-2001) 清华(2001-date)

博文

Recent Progresses in Pattern Recognition 精选

已有 2940 次阅读 2019-4-12 22:03 |个人分类:S and T|系统分类:海外观察

For new readers and those who request to be “好友 good friends” please read my 公告 first

 

I have written earlier about Pattern Recognition and Deep Learning

Decision Making,Backward Propagation, and Deep Learning http://blog.sciencenet.cn/blog-1565-910916.html which dealt with computational issues of multilayered neural networks.  Today I attended a talk on deep learning with the following title and abstract.

 

Richard Zemel of the University of Toronto and the Vector Institute for Artificial Intelligence will give a talk entitled

 

“Controlling the Black Box: Learning Manipulable and Fair Representations”

 

Thursday, April 11, 2019

3:00 p.m.

Maxwell Dworkin G115

  

Machine learning models, and more specifically deep neural networks, are achieving state-of-the-art performance on difficult pattern-recognition tasks such as object recognition, speech recognition, drug discovery, and more. However, deep networks are notoriously difficult to understand, both in how they arrive at and how to affect their responses. As these systems become more prevalent in real-world applications it is essential to allow users to exert more control over the learning system. In particular a wide range of applications can be facilitated by exerting some structure over the learned representations, to enable users to manipulate, interpret, and in some cases obfuscate the representations. In this talk I will discuss recent work that makes some steps towards these goals, allowing users to interact with and control representations.

 

Speaker:  Richard Zemel is a Professor of Computer Science and Industrial Research Chair in Machine Learning at the University of Toronto, and a co-founder and the Research Director at the Vector Institute for Artificial Intelligence.  Prior to that he was on the faculty at the University of Arizona, and a Postdoctoral Fellow at the Salk Institute and at CMU. He received the B.Sc. in History & Science from Harvard, and a Ph.D. in Computer Science from the University of Toronto. His awards and honors include a Young Investigator Award from the ONR and a US Presidential Scholar award.  He is a Senior Fellow of the Canadian Institute for Advanced Research, an NVIDIA Pioneer of AI, and a member of the NeurIPS Advisory Board.


Reading the abstract would not tell an outsider like me what this is all about. The following is my takeaway about what computer scientists like Zemel are working on.

A multilayered network takes in a set of visual pixel readings and through layers of weighted connections to produce at the output a decision (e.g., the pixels contain a recognized person’s face). However, nowadays researchers are also interested in additional hidden or latent representations in the intermediate layers. For example, does the face show emotion, happy or sad? Beard or no beard? Sex and gender? And other attributes that may be useful for advertisers. Can these latent representations be identified and made generally useful in other classification problems. The general approaches to these solved or unsolved questions rely on prior and posterior probability distributions and the concept of Invertible networks. The work is interesting, challenging, and difficult.

Note added 4/16/2019. For an elementary introduction to pattern recognition see http://blog.sciencenet.cn/blog-1565-243352.html 



http://blog.sciencenet.cn/blog-1565-1172945.html

上一篇:Finger Print vs. DNA Evidence in Solving Crimes
下一篇:[转载]How to raise successful children

3 郑永军 许培扬 李天成

该博文允许注册用户评论 请点击登录 评论 (0 个评论)

数据加载中...

Archiver|手机版|科学网 ( 京ICP备14006957 )

GMT+8, 2019-10-24 00:22

Powered by ScienceNet.cn

Copyright © 2007- 中国科学报社

返回顶部