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 2011-06-21 14:15
Analysis and Improvement of Policy Gradient Estimation
Tingting Zhao, Hirotaka Hachiya, Gang Niu, Masashi Sugiyama (Tokyo Inst. of Tech.) IBISML2011-12
Abstract (in Japanese) (See Japanese page) 
(in English) Policy gradient is a useful model-free reinforcement learning approach,
but it tends to suffer from instability of gradient estimates.
In this paper, we analyze and improve the stability of policy gradient methods. We first prove that the variance of gradient estimates in the PGPE (policy gradients with parameter-based exploration) method is smaller than that of the classical REINFORCE method under a mild assumption. We then derive the optimal baseline for PGPE, which contributes to further reducing the variance. We also theoretically show that PGPE with the optimal baseline is more preferable than REINFORCE with the optimal baseline in terms of the variance of gradient estimates.
Finally, we demonstrate the usefulness of the improved PGPE method through experiments.
Keyword (in Japanese) (See Japanese page) 
(in English) Reinforcement Learning / Policy gradients / Optimal Baseline / / / / /  
Reference Info. IEICE Tech. Rep., vol. 111, no. 87, IBISML2011-12, pp. 83-89, June 2011.
Paper # IBISML2011-12 
Date of Issue 2011-06-13 (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 IBISML2011-12

Conference Information
Committee IBISML  
Conference Date 2011-06-20 - 2011-06-21 
Place (in Japanese) (See Japanese page) 
Place (in English) Takeda Hall 
Topics (in Japanese) (See Japanese page) 
Topics (in English) Machine learning and its applications 
Paper Information
Registration To IBISML 
Conference Code 2011-06-IBISML 
Language English 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Analysis and Improvement of Policy Gradient Estimation 
Sub Title (in English)  
Keyword(1) Reinforcement Learning  
Keyword(2) Policy gradients  
Keyword(3) Optimal Baseline  
Keyword(4)  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Tingting Zhao  
1st Author's Affiliation Tokyo Institute of Technology (Tokyo Inst. of Tech.)
2nd Author's Name Hirotaka Hachiya  
2nd Author's Affiliation Tokyo Institute of Technology (Tokyo Inst. of Tech.)
3rd Author's Name Gang Niu  
3rd Author's Affiliation Tokyo Institute of Technology (Tokyo Inst. of Tech.)
4th Author's Name Masashi Sugiyama  
4th Author's Affiliation Tokyo Institute of Technology (Tokyo Inst. of Tech.)
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 2011-06-21 14:15:00 
Presentation Time 30 minutes 
Registration for IBISML 
Paper # IBISML2011-12 
Volume (vol) vol.111 
Number (no) no.87 
Page pp.83-89 
#Pages
Date of Issue 2011-06-13 (IBISML) 


[Return to Top Page]

[Return to IEICE Web Page]


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