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 2013-07-19 14:30
Statistical Mechanics of node-perturbation Learning using two independent noises
Kazuyuki Hara (Nihon Univ.), Kentaro Katahira, Masato Okada (Univ. of Tokyo) NC2013-17
Abstract (in Japanese) (See Japanese page) 
(in English) Node perturbation learning is a stochastic gradient descent method for neural networks. It estimates the gradient by comparing an evaluation of the perturbed output and the unperturbed output performance, which we call the baseline. However, the unperturbed performance is a difficult setting. Cho et al. proposed node-perturbation using two independent noises, and it is tractable for real problem. In this report, we analyze the node-perturbation learning using two independent noises by using statistical mechanics methods. From the results, we find that the
residual error becomes small compared to the original one when the variance of two noises is the same.
Keyword (in Japanese) (See Japanese page) 
(in English) node-perturbation learning / two independent noises / on-line learning / generalization error / statistical mechanics method / / /  
Reference Info. IEICE Tech. Rep., vol. 113, no. 148, NC2013-17, pp. 13-18, July 2013.
Paper # NC2013-17 
Date of Issue 2013-07-12 (NC) 
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 NC2013-17

Conference Information
Committee NC MBE  
Conference Date 2013-07-19 - 2013-07-20 
Place (in Japanese) (See Japanese page) 
Place (in English) The University of Tokushima 
Topics (in Japanese) (See Japanese page) 
Topics (in English) ME, general 
Paper Information
Registration To NC 
Conference Code 2013-07-NC-MBE 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Statistical Mechanics of node-perturbation Learning using two independent noises 
Sub Title (in English)  
Keyword(1) node-perturbation learning  
Keyword(2) two independent noises  
Keyword(3) on-line learning  
Keyword(4) generalization error  
Keyword(5) statistical mechanics method  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Kazuyuki Hara  
1st Author's Affiliation Nihon University (Nihon Univ.)
2nd Author's Name Kentaro Katahira  
2nd Author's Affiliation The University of Tokyo (Univ. of Tokyo)
3rd Author's Name Masato Okada  
3rd Author's Affiliation The University of Tokyo (Univ. of Tokyo)
4th Author's Name  
4th Author's Affiliation ()
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 2013-07-19 14:30:00 
Presentation Time 25 minutes 
Registration for NC 
Paper # NC2013-17 
Volume (vol) vol.113 
Number (no) no.148 
Page pp.13-18 
#Pages
Date of Issue 2013-07-12 (NC) 


[Return to Top Page]

[Return to IEICE Web Page]


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