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 2013-11-13 15:45
[Poster Presentation] Sample Complexity Reduction in Reinforcement Learning by Transferred Transition and Reward Probability
Kouta Oguni, Kazuyuki Narisawa, Ayumi Shinohara (Tohoku Univ.) IBISML2013-54
Abstract (in Japanese) (See Japanese page) 
(in English) Most existing reinforcement learning algorithms are not very efficient in real environmental problems. Because, they have to try many times till they get an optimal policy. In this paper, we apply transfer learning to reinforcement learning for efficient learning. We propose a new algorithm called TR-MAX. The algorithm transfers transition and reward probabilities from a source task to a target task. We show that the sample complexity of TR-MAX is smaller than that of the base algorithm. Finally, we show that the performance of our algorithm is better than that of the base algorithm in a maze task.
Keyword (in Japanese) (See Japanese page) 
(in English) Reinforcement Learning / Transfer Learning / Sample Complexity / PAC-MDP / / / /  
Reference Info. IEICE Tech. Rep., vol. 113, no. 286, IBISML2013-54, pp. 139-146, Nov. 2013.
Paper # IBISML2013-54 
Date of Issue 2013-11-05 (IBISML) 
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 IBISML2013-54

Conference Information
Committee IBISML  
Conference Date 2013-11-10 - 2013-11-13 
Place (in Japanese) (See Japanese page) 
Place (in English) Tokyo Institute of Technology, Kuramae-Kaikan 
Topics (in Japanese) (See Japanese page) 
Topics (in English) The 16th IBIS Workshop & The 2nd IBIS Tutorial 
Paper Information
Registration To IBISML 
Conference Code 2013-11-IBISML 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Sample Complexity Reduction in Reinforcement Learning by Transferred Transition and Reward Probability 
Sub Title (in English)  
Keyword(1) Reinforcement Learning  
Keyword(2) Transfer Learning  
Keyword(3) Sample Complexity  
Keyword(4) PAC-MDP  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Kouta Oguni  
1st Author's Affiliation Tohoku University (Tohoku Univ.)
2nd Author's Name Kazuyuki Narisawa  
2nd Author's Affiliation Tohoku University (Tohoku Univ.)
3rd Author's Name Ayumi Shinohara  
3rd Author's Affiliation Tohoku University (Tohoku Univ.)
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 2013-11-13 15:45:00 
Presentation Time 180 minutes 
Registration for IBISML 
Paper # IBISML2013-54 
Volume (vol) vol.113 
Number (no) no.286 
Page pp.139-146 
#Pages
Date of Issue 2013-11-05 (IBISML) 


[Return to Top Page]

[Return to IEICE Web Page]


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