Presentation 2004/6/18
Efficient learning of real-time motor skills by parallel policies
BISSMARCK Fredrik, Hiroyuki NAKAHARA, Kenji DOYA, Okihide HIKOSAKA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Motor control depends on modalities with different latencies. In this paper, we consider how to combine different sensory modalities for real-time optimal movement. We suggest that the modality with shorter latency should have an advantage. We demonstrate with an example simulation of a visuomotor sequence reaching task. In our example, the behaviour is initially a visually guided, point-to-point movement. By learning, the sensorimotor modality gradually gains influence, and the movement became faster and optimized to the full sequence. After learning, the sensorimotor modality were dominant in executing the sequence, while the visual module still played a vital part when the arm lost control of the execution. We found that the shorter the relative latency of the sensorimotor modality, the more influential it became of the acquired skill.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) sequence learning / reinforcement learning / basal ganglia / motor control
Paper # NC2004-32
Date of Issue

Conference Information
Committee NC
Conference Date 2004/6/18(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 ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Efficient learning of real-time motor skills by parallel policies
Sub Title (in English)
Keyword(1) sequence learning
Keyword(2) reinforcement learning
Keyword(3) basal ganglia
Keyword(4) motor control
1st Author's Name BISSMARCK Fredrik
1st Author's Affiliation Nara Institute of Science and Technology:Computational Neuroscience Labs, ATR International:Creating the Brain, CREST, Japan Science and Technology Corporation()
2nd Author's Name Hiroyuki NAKAHARA
2nd Author's Affiliation Laboratory for Mathematical Neuroscience, RIKEN Brain Science Institute
3rd Author's Name Kenji DOYA
3rd Author's Affiliation Nara Institute of Science and Technology:Computational Neuroscience Labs, ATR International:Creating the Brain, CREST, Japan Science and Technology Corporation
4th Author's Name Okihide HIKOSAKA
4th Author's Affiliation Laboratory of Sensorimotor Research, National Eye Institute, National Institute of Health
Date 2004/6/18
Paper # NC2004-32
Volume (vol) vol.104
Number (no) 140
Page pp.pp.-
#Pages 6
Date of Issue