Skip navigation
Skip navigation

Asymptotic learnability of reinforcement problems with arbitrary dependence

Ryabko, Daniil; Hutter, Marcus

Description

We address the problem of reinforcement learning in which observations may exhibit an arbitrary form of stochastic dependence on past observations and actions, i.e. environments more general than (PO) MDPs. The task for an agent is to attain the best possible asymptotic reward where the true generating environment is unknown but belongs to a known countable family of environments. We find some sufficient conditions on the class of environments under which an agent exists which attains the best...[Show more]

dc.contributor.authorRyabko, Daniil
dc.contributor.authorHutter, Marcus
dc.coverage.spatialBarcelona Spain
dc.date.accessioned2015-12-07T22:55:41Z
dc.date.createdOctober 7-10 2006
dc.identifier.isbn3540466495
dc.identifier.urihttp://hdl.handle.net/1885/28514
dc.description.abstractWe address the problem of reinforcement learning in which observations may exhibit an arbitrary form of stochastic dependence on past observations and actions, i.e. environments more general than (PO) MDPs. The task for an agent is to attain the best possible asymptotic reward where the true generating environment is unknown but belongs to a known countable family of environments. We find some sufficient conditions on the class of environments under which an agent exists which attains the best asymptotic reward for any environment in the class. We analyze how tight these conditions are and how they relate to different probabilistic assumptions known in reinforcement learning and related fields, such as Markov Decision Processes and mixing conditions.
dc.publisherSpringer
dc.relation.ispartofseriesInternational Conference on Algorithmic Learning Theory (ALT 2006)
dc.sourceProceedings of International Conference on Algorithmic Learning Theory (ALT 2006)
dc.subjectKeywords: Asymptotic stability; Decision theory; Intelligent agents; Markov processes; Problem solving; Asymptotic learnability; Markov Decision Processes (MDP); Reinforcement learning; Learning systems
dc.titleAsymptotic learnability of reinforcement problems with arbitrary dependence
dc.typeConference paper
local.description.notesImported from ARIES
local.description.refereedYes
dc.date.issued2006
local.identifier.absfor080401 - Coding and Information Theory
local.identifier.absfor010405 - Statistical Theory
local.identifier.absfor080199 - Artificial Intelligence and Image Processing not elsewhere classified
local.identifier.ariespublicationu8803936xPUB58
local.type.statusPublished Version
local.contributor.affiliationRyabko, Daniil, IDSIA-Istituto Dalle Molle di Studi sull Intelligenza Artificiale
local.contributor.affiliationHutter, Marcus, College of Engineering and Computer Science, ANU
local.description.embargo2037-12-31
local.bibliographicCitation.startpage334
local.bibliographicCitation.lastpage347
dc.date.updated2015-12-07T13:00:53Z
local.identifier.scopusID2-s2.0-33750698276
CollectionsANU Research Publications



Items in Open Research are protected by copyright, with all rights reserved, unless otherwise indicated.

Updated:  19 May 2020/ Responsible Officer:  University Librarian/ Page Contact:  Library Systems & Web Coordinator