Presentation 2011-01-24
Model Adaptation with Bayesian Hierarchical Models
Hideki ASOH,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Model adaptation is a process of modifying general model which is trained with large amount of training data to adapt a specific task/user using small amount of adaptation data regarding the task/user. Typical examples are acoustic model adaptation and language model adaptation for speech recognition systems. Model adaptation is a kind of transfer learning and multi-task learning. Bayesian hierarchical modeling is known as a general tool for multi-task learning and widely used in various areas such as marketing, ecology , medicine, education, etc. to model the heterogeneity of the phenomena. In this work, a model adaptation procedure using Bayesian hierarchical models is given and applied to the problem of preference modeling, where a model trained with large amount of virtual situation data is adapted to real situation. Experimental results with context-aware food preference data demonstrate effectiveness of the proposed method.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Bayesian hierarchical model / model adaptation / transfer learning / preference model
Paper # NLP2010-140,NC2010-104
Date of Issue

Conference Information
Committee NC
Conference Date 2011/1/17(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) Model Adaptation with Bayesian Hierarchical Models
Sub Title (in English)
Keyword(1) Bayesian hierarchical model
Keyword(2) model adaptation
Keyword(3) transfer learning
Keyword(4) preference model
1st Author's Name Hideki ASOH
1st Author's Affiliation National Institute of Advanced Industrial Science and Technology (AIST)()
Date 2011-01-24
Paper # NLP2010-140,NC2010-104
Volume (vol) vol.110
Number (no) 388
Page pp.pp.-
#Pages 6
Date of Issue