Presentation 2007-10-18
Variational Bayesian Hidden Markov Models for extracting spatiotemporal spike pattern
Kentaro KATAHIRA, Jun NISHIKAWA, Kazuo OKANOYA, Masato OKADA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Hidden Markov Model (HMM) is used to extracting spatio-temporal pattern from spikes recorded by multielectrode. The EM algorithm that archieves maximum likelihood estimation is used to train HMM in that context. Therefore the previous method has problems of overlearning, and the number of hidden states must be determined by trail-and-error. To overcome those problems, we applied Variational Byases (VB) method to train the model. Our method can determine the number of states automatically and avoid overlearning. We performed an experiment on synthetic data and a real neural data from nucleus HVC of Bengalese finch during listening to own song.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Maltielectrode spike trains / Variational Bayes / Hidden Markov Model / HVC
Paper # NC2007-34
Date of Issue

Conference Information
Committee NC
Conference Date 2007/10/11(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) Variational Bayesian Hidden Markov Models for extracting spatiotemporal spike pattern
Sub Title (in English)
Keyword(1) Maltielectrode spike trains
Keyword(2) Variational Bayes
Keyword(3) Hidden Markov Model
Keyword(4) HVC
1st Author's Name Kentaro KATAHIRA
1st Author's Affiliation Graduate School of Frontier Sciences, The University of Tokyo:RIKEN Brain Science Institute()
2nd Author's Name Jun NISHIKAWA
2nd Author's Affiliation RIKEN Brain Science Institute
3rd Author's Name Kazuo OKANOYA
3rd Author's Affiliation RIKEN Brain Science Institute
4th Author's Name Masato OKADA
4th Author's Affiliation Graduate School of Frontier Sciences, The University of Tokyo:RIKEN Brain Science Institute
Date 2007-10-18
Paper # NC2007-34
Volume (vol) vol.107
Number (no) 263
Page pp.pp.-
#Pages 6
Date of Issue