Simultaneous abstract and concrete reinforcement learning

Tiago Matos; Yannick P. Bergamo; Valdinei F. Da Silva; Fabio G. Cozman; Anna H. Reali Costa

Simultaneous abstract and concrete reinforcement learning

Tiago Matos, Yannick P. Bergamo, Valdinei F. Da Silva, Fabio G. Cozman, Anna H. Reali Costa

Research output: Contribution to conference › Conference Paper

6 Scopus citations

Abstract

Suppose an agent builds a policy that satisfactorily solves a decision problem; suppose further that some aspects of this policy are abstracted and used as starting point in a new, different decision problem. How can the agent accrue the benefits of the abstract policy in the new concrete problem? In this paper we propose a framework for simultaneous reinforcement learning, where the abstract policy helps start up the policy for the concrete problem, and both policies are refined through exploration. We report experiments that demonstrate that our framework is effective in speeding up policy construction for practical problems. Copyright © 2011, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.

Original language	American English
Pages	82-89
Number of pages	8
State	Published - 1 Dec 2011
Externally published	Yes
Event	SARA 2011 - Proceedings of the 9th Symposium on Abstraction, Reformulation, and Approximation - Duration: 1 Dec 2011 → …

Conference

Conference	SARA 2011 - Proceedings of the 9th Symposium on Abstraction, Reformulation, and Approximation
Period	1/12/11 → …

Cite this

@conference{083cb3c706594612a9aade20561ddcc5,

title = "Simultaneous abstract and concrete reinforcement learning",

abstract = "Suppose an agent builds a policy that satisfactorily solves a decision problem; suppose further that some aspects of this policy are abstracted and used as starting point in a new, different decision problem. How can the agent accrue the benefits of the abstract policy in the new concrete problem? In this paper we propose a framework for simultaneous reinforcement learning, where the abstract policy helps start up the policy for the concrete problem, and both policies are refined through exploration. We report experiments that demonstrate that our framework is effective in speeding up policy construction for practical problems. Copyright {\textcopyright} 2011, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.",

author = "Tiago Matos and Bergamo, {Yannick P.} and {Da Silva}, {Valdinei F.} and Cozman, {Fabio G.} and {Reali Costa}, {Anna H.}",

year = "2011",

month = dec,

day = "1",

language = "American English",

pages = "82--89",

note = "SARA 2011 - Proceedings of the 9th Symposium on Abstraction, Reformulation, and Approximation ; Conference date: 01-12-2011",

}

TY - CONF

T1 - Simultaneous abstract and concrete reinforcement learning

AU - Matos, Tiago

AU - Bergamo, Yannick P.

AU - Da Silva, Valdinei F.

AU - Cozman, Fabio G.

AU - Reali Costa, Anna H.

PY - 2011/12/1

Y1 - 2011/12/1

N2 - Suppose an agent builds a policy that satisfactorily solves a decision problem; suppose further that some aspects of this policy are abstracted and used as starting point in a new, different decision problem. How can the agent accrue the benefits of the abstract policy in the new concrete problem? In this paper we propose a framework for simultaneous reinforcement learning, where the abstract policy helps start up the policy for the concrete problem, and both policies are refined through exploration. We report experiments that demonstrate that our framework is effective in speeding up policy construction for practical problems. Copyright © 2011, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.

AB - Suppose an agent builds a policy that satisfactorily solves a decision problem; suppose further that some aspects of this policy are abstracted and used as starting point in a new, different decision problem. How can the agent accrue the benefits of the abstract policy in the new concrete problem? In this paper we propose a framework for simultaneous reinforcement learning, where the abstract policy helps start up the policy for the concrete problem, and both policies are refined through exploration. We report experiments that demonstrate that our framework is effective in speeding up policy construction for practical problems. Copyright © 2011, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.

UR - https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=84868005896&origin=inward

UR - https://www.scopus.com/inward/citedby.uri?partnerID=HzOxMe3b&scp=84868005896&origin=inward

M3 - Conference Paper

SP - 82

EP - 89

T2 - SARA 2011 - Proceedings of the 9th Symposium on Abstraction, Reformulation, and Approximation

Y2 - 1 December 2011

ER -

Simultaneous abstract and concrete reinforcement learning

Abstract

Conference

Other files and links

Fingerprint

Cite this