IEICE Technical Committee Submission System
Conference Paper's Information
Online Proceedings
[Sign in]
Tech. Rep. Archives
 Go Top Page Go Previous   [Japanese] / [English] 

Paper Abstract and Keywords
Presentation 2018-06-10 10:45
Use of recurrent infomax to improve the information processing capability of input-driven recurrent neural networks
Hisashi Iwade (Kyoto Univ.), Kohei Nakajima (Univ. Tokyo/JST PRESTO), Takuma Tanaka (Shiga Univ.), Toshio Aoyagi (Kyoto Univ.) NLP2018-42 CCS2018-15
Abstract (in Japanese) (See Japanese page) 
(in English) Reservoir computing (RC) is known to be a novel information processing framework, which exploits the inherent transient dynamics of recurrent neural networks (RNNs) as a computational resource. RC possesses high information processing capability but, in some cases, the empirical optimization using the control parameters is required. In this study, we propose recurrent infomax (RI) as a alternative method to the empirical optimization. RI is an unsupervised learning scheme that maximizes the mutual information of RNNs by adjusting the connection strengths of the network. Concequently, a delay-line structure can be observed on the network optimized by RI and the network possesses superior short-term memory, which is the ability to encode the information of the input stream in its transient dynamics.
Keyword (in Japanese) (See Japanese page) 
(in English) recurrent neural network / recurrent infomax / / / / / /  
Reference Info. IEICE Tech. Rep., vol. 118, no. 76, CCS2018-15, pp. 79-84, June 2018.
Paper # CCS2018-15 
Date of Issue 2018-06-01 (NLP, CCS) 
ISSN Print edition: ISSN 0913-5685    Online edition: ISSN 2432-6380
Copyright
and
reproduction
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034)
Download PDF NLP2018-42 CCS2018-15

Conference Information
Committee NLP CCS  
Conference Date 2018-06-08 - 2018-06-10 
Place (in Japanese) (See Japanese page) 
Place (in English) Kyoto Terrsa 
Topics (in Japanese) (See Japanese page) 
Topics (in English) Synchronization, Networks, etc 
Paper Information
Registration To CCS 
Conference Code 2018-06-NLP-CCS 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Use of recurrent infomax to improve the information processing capability of input-driven recurrent neural networks 
Sub Title (in English)  
Keyword(1) recurrent neural network  
Keyword(2) recurrent infomax  
Keyword(3)  
Keyword(4)  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Hisashi Iwade  
1st Author's Affiliation Kyoto University (Kyoto Univ.)
2nd Author's Name Kohei Nakajima  
2nd Author's Affiliation University of Tokyo/JST PRESTO (Univ. Tokyo/JST PRESTO)
3rd Author's Name Takuma Tanaka  
3rd Author's Affiliation Shiga University (Shiga Univ.)
4th Author's Name Toshio Aoyagi  
4th Author's Affiliation Kyoto University (Kyoto Univ.)
5th Author's Name  
5th Author's Affiliation ()
6th Author's Name  
6th Author's Affiliation ()
7th Author's Name  
7th Author's Affiliation ()
8th Author's Name  
8th Author's Affiliation ()
9th Author's Name  
9th Author's Affiliation ()
10th Author's Name  
10th Author's Affiliation ()
11th Author's Name  
11th Author's Affiliation ()
12th Author's Name  
12th Author's Affiliation ()
13th Author's Name  
13th Author's Affiliation ()
14th Author's Name  
14th Author's Affiliation ()
15th Author's Name  
15th Author's Affiliation ()
16th Author's Name  
16th Author's Affiliation ()
17th Author's Name  
17th Author's Affiliation ()
18th Author's Name  
18th Author's Affiliation ()
19th Author's Name  
19th Author's Affiliation ()
20th Author's Name  
20th Author's Affiliation ()
Speaker Author-1 
Date Time 2018-06-10 10:45:00 
Presentation Time 25 minutes 
Registration for CCS 
Paper # NLP2018-42, CCS2018-15 
Volume (vol) vol.118 
Number (no) no.75(NLP), no.76(CCS) 
Page pp.79-84 
#Pages
Date of Issue 2018-06-01 (NLP, CCS) 


[Return to Top Page]

[Return to IEICE Web Page]


The Institute of Electronics, Information and Communication Engineers (IEICE), Japan