推荐星级:
  • 1
  • 2
  • 3
  • 4
  • 5

神经网络与深度学习综述DeepLearning

更新时间:2019-11-13 10:54:16 大小:266K 上传用户:xuzhen1查看TA发布的资源 标签:神经网络深度deep learning 下载积分:0分 评价赚积分 (如何评价?) 打赏 收藏 评论(0) 举报

资料介绍

Preface This is the draft of an invited Deep Learning(DL) overview. One of its goals is to assign credit to those who contributed to the present state of the art.I acknowledge the limitations of attempting to achieve this goal. The DL research community itself may be viewed as a continually evolving, deep network of scientists who have in? uenced each other in complex ways. Starting from recent DL results,I tried to traceback the origins of relevant ideas through the past half century and beyond, sometimes using "locasearch "

to follow citations of citations backwards in time. Since not all DL publications properly acknowledge earlier relevant work, additional global search strategies were employed, aided by consulting numerous neural network experts. As a result, the present draft mostly consists of references(about 800 entries so far). Nevertheless, through an expert selection bias I may have missed important work.A related bias was surely introduced by my special familiarity with the work of my own DL research group in the past quarter-century. For these reasons, the present draft should be viewed as merely a snapshot of an ongoing credit assignment process. To help improve it, please do not hesitate to send corrections and suggestions to juergen@ idsia. ch.


部分文件列表

文件名 大小
神经网络与深度学习综述DeepLearning.pdf 266K

部分页面预览

(完整内容请下载后查看)
Draft: Deep Learning in Neural Networks: An Overview  
Technical Report IDSIA-03-14 / arXiv:1404.7828 (v1.5) [cs.NE]  
Ju¨rgen Schmidhuber  
The Swiss AI Lab IDSIA  
Istituto Dalle Molle di Studi sull IntelligenzAarti?ciale  
University of Lugano & SUPSI  
Galleria 2, 6928 Manno-Lugano  
Switzerland  
15 May 2014  
Abstract  
In recent years, deep arti?cial neural networks (including recurrent ones) have won numerous con-  
testsin pattern recognition and machine learning. This historical surveycompactly summarisesrelevant  
work, much of it from the previous millennium. Shallow and deep learners are distinguished by the  
depth of their credit assignment paths, which are chains of possibly learnable, causal links between ac-  
tions and effects. I review deep supervised learning (also recapitulating the history of backpropagation),  
unsupervisedlearning, reinforcement learning & evolutionary computation, and indirect searchfor short  
programs encoding deepand large networks.  
juergen/DeepLearning30April2014.pdf  
juergen/DeepLearning30April2014.tex  
PDF of earlier draft (v1): http://www.idsia.ch/  
LATEX source: 
juergen/bib.bib  
Complete BIBTEX ?le: http://www.idsia.ch/  
Preface  
This is the draft of an invited Deep Learning (DL) overview. One of its goals is to assign credit to those  
who contributed to the present state of the art. I acknowledge the limitations of attempting to achieve  
this goal. The DL research community itself may be viewed as a continually evolving, deep network of  
scientists who have in?uenced each other in complex ways. Starting from recent DL results, I tried to trace  
back the origins of relevant ideas through the past half century and beyond, sometimes using locasl earch ”  
to follow citations of citations backwards in time. Since not all DL publications properly acknowledge  
earlier relevant work, additional global search strategies were employed, aided by consulting numerous  
neural network experts. As a result, the present draft mostly consists of references (about 800 entries so  
far). Nevertheless, through an expert selection bias I may have missed important work. A related bias  
was surely introduced by my special familiarity with the work of my own DL research group in the past  
quarter-century. For these reasons, the present draft should be viewed as merely a snapshot of an ongoing  
credit assignment process. To help improve it, please do not hesitate to send corrections and suggestions to  
.  
1

全部评论(0)

暂无评论

上传资源 上传优质资源有赏金

  • 打赏
  • 30日榜单

推荐下载