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 2019-03-05 14:30
Efficient Exploration by Variational Information Maximizing Exploration on Reinforcement Learning
Kazuki Doi, Keigo Okawa (Gifu Univ.), Motoki Shiga (Gifu Univ./JST/RIKEN) IBISML2018-107
Abstract (in Japanese) (See Japanese page) 
(in English) In reinforcement learning,the policy function may not be optimized properly if the observed state space is limited to local sub-space.To facilitate exploration of unseen states, Variational Information Maximizing Exploration (VIME), which evaluates entropy of the next action and state, is useful. We propose to combine VIME and Trust Region Policy Optimization (TRPO). The performance of our method was evaluated using three reinforcement learning tasks. The results shows that VIME drastically improves the acquisition reward for the task that needs a lot of steps to achieve the goal state.
Keyword (in Japanese) (See Japanese page) 
(in English) Reinforcement Learning / Variational Information Maximizing Exploration / Trust Region Policy Optimization / / / / /  
Reference Info. IEICE Tech. Rep., vol. 118, no. 472, IBISML2018-107, pp. 17-22, March 2019.
Paper # IBISML2018-107 
Date of Issue 2019-02-26 (IBISML) 
ISSN Print edition: ISSN 0913-5685  Online edition: ISSN 2432-6380
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 IBISML2018-107

Conference Information
Committee IBISML  
Conference Date 2019-03-05 - 2019-03-06 
Place (in Japanese) (See Japanese page) 
Place (in English) RIKEN AIP 
Topics (in Japanese) (See Japanese page) 
Topics (in English) Machine learning, etc. 
Paper Information
Registration To IBISML 
Conference Code 2019-03-IBISML 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Efficient Exploration by Variational Information Maximizing Exploration on Reinforcement Learning 
Sub Title (in English)  
Keyword(1) Reinforcement Learning  
Keyword(2) Variational Information Maximizing Exploration  
Keyword(3) Trust Region Policy Optimization  
1st Author's Name Kazuki Doi  
1st Author's Affiliation Gifu University (Gifu Univ.)
2nd Author's Name Keigo Okawa  
2nd Author's Affiliation Gifu University (Gifu Univ.)
3rd Author's Name Motoki Shiga  
3rd Author's Affiliation Gifu University/Japan Science and Technology Agency/RIKEN (Gifu Univ./JST/RIKEN)
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 ()
Date Time 2019-03-05 14:30:00 
Presentation Time 30 
Registration for IBISML 
Paper # IEICE-IBISML2018-107 
Volume (vol) IEICE-118 
Number (no) no.472 
Page pp.17-22 
#Pages IEICE-6 
Date of Issue IEICE-IBISML-2019-02-26 

[Return to Top Page]

[Return to IEICE Web Page]

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