Presentation 2007-03-14
Action-Oriented State Coding by Neighbourhood Component Analysis
Makoto OTSUKA, Eiji UCHIBE, Kenji DOYA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) The performance of reinforcement learning severely depends on its underlying state representation; therefore, the automatic acquisition of a task-dependent state space is a major topic in the field of reinforcement learirig. This research proposes a novel way to construct an efficient and task-dependent state representation by integrating two methods: the neighbourhood component analysis (NCA) and the instance-based reinforcement learning (IBRL). In three different simulation experiments, the performance of different dimensionality reduction techniques are compared with the proposed method. The results of the experiments show that the proposed method finds important features and constructs an effective task-dependent state representation automatically.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) neighbourhood component analysis / reinforcement learning / stochastic nearest neighbour / distance metric / state representation / dimensionality reduction
Paper # NC2006-149
Date of Issue

Conference Information
Committee NC
Conference Date 2007/3/7(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Action-Oriented State Coding by Neighbourhood Component Analysis
Sub Title (in English)
Keyword(1) neighbourhood component analysis
Keyword(2) reinforcement learning
Keyword(3) stochastic nearest neighbour
Keyword(4) distance metric
Keyword(5) state representation
Keyword(6) dimensionality reduction
1st Author's Name Makoto OTSUKA
1st Author's Affiliation Initial Research Project, Okinawa Institute of Science and Technology:Nara Institute of Science and Technology()
2nd Author's Name Eiji UCHIBE
2nd Author's Affiliation Initial Research Project, Okinawa Institute of Science and Technology
3rd Author's Name Kenji DOYA
3rd Author's Affiliation Initial Research Project, Okinawa Institute of Science and Technology:Nara Institute of Science and Technology
Date 2007-03-14
Paper # NC2006-149
Volume (vol) vol.106
Number (no) 588
Page pp.pp.-
#Pages 6
Date of Issue