Paper Abstract and Keywords |
Presentation |
2016-11-17 14:00
Incremental Natural Actor Critic with Importance Weight Aware Update Ryo Iwaki (Osaka Univ.), Hiroki Yokoyama (Tamagawa Univ.), Minoru Asada (Osaka Univ.) IBISML2016-81 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Appropriate tuning of step-size parameter is crucial for reinforcement learning, as well as other machine learning techniques. For linear regression cases, importance weight aware update (IWAU) guarantees a safety; the updated parameter will not overshoot the stationary point of loss function. If a squared loss is utilized, IWAU can be derived as a simple closed form. In this paper, We focus on the natural policy gradient (NPG) methods, which can be derived as linear function approximation and squared loss, and derive new adaptive step-size for NPG estimation method, which is safe and free from step-size meta-parameter tuning. We confirm the usefulness of the proposed step-size in classical benchmark for natural policy gradient method. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
reinforcement learning / natural policy gradient method / adaptive step-size / incremental learning / importance weight aware update / / / |
Reference Info. |
IEICE Tech. Rep., vol. 116, no. 300, IBISML2016-81, pp. 251-257, Nov. 2016. |
Paper # |
IBISML2016-81 |
Date of Issue |
2016-11-09 (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 |
IBISML2016-81 |
Conference Information |
Committee |
IBISML |
Conference Date |
2016-11-16 - 2016-11-18 |
Place (in Japanese) |
(See Japanese page) |
Place (in English) |
Kyoto Univ. |
Topics (in Japanese) |
(See Japanese page) |
Topics (in English) |
Information-Based Induction Science Workshop (IBIS2016) |
Paper Information |
Registration To |
IBISML |
Conference Code |
2016-11-IBISML |
Language |
Japanese |
Title (in Japanese) |
(See Japanese page) |
Sub Title (in Japanese) |
(See Japanese page) |
Title (in English) |
Incremental Natural Actor Critic with Importance Weight Aware Update |
Sub Title (in English) |
|
Keyword(1) |
reinforcement learning |
Keyword(2) |
natural policy gradient method |
Keyword(3) |
adaptive step-size |
Keyword(4) |
incremental learning |
Keyword(5) |
importance weight aware update |
Keyword(6) |
|
Keyword(7) |
|
Keyword(8) |
|
1st Author's Name |
Ryo Iwaki |
1st Author's Affiliation |
Osaka University (Osaka Univ.) |
2nd Author's Name |
Hiroki Yokoyama |
2nd Author's Affiliation |
Tamagawa University (Tamagawa Univ.) |
3rd Author's Name |
Minoru Asada |
3rd Author's Affiliation |
Osaka University (Osaka 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 |
2016-11-17 14:00:00 |
Presentation Time |
180 minutes |
Registration for |
IBISML |
Paper # |
IBISML2016-81 |
Volume (vol) |
vol.116 |
Number (no) |
no.300 |
Page |
pp.251-257 |
#Pages |
7 |
Date of Issue |
2016-11-09 (IBISML) |
|