Presentation 2002/3/11
Bayes Generalization Errors of Reduced Rank Approximation
Kazuho WATANABE, Sumlo WATANABE,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Reduced rank approximation corresponds to a three-layer linear neural network with reduced hidden layer size. A lot of attention is paid to the properties of it and some basic questions can sometimes be answered analytically for its linearity. However, as is the case with other layered models such as neural networks and gaussian mixtures, its true parameter is not identifiable. Therefore some problems related to learning remain unsolved. In this paper, we analyze the generalization error of the reduced rank approximation in Bayesian estimation, and figure out its upper bound using an algebraic geometrical method.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Reduced rank approximation / Bayesian estimation / generalization error / algebraic geometry / singularities
Paper # NC2001-149
Date of Issue

Conference Information
Committee NC
Conference Date 2002/3/11(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) Bayes Generalization Errors of Reduced Rank Approximation
Sub Title (in English)
Keyword(1) Reduced rank approximation
Keyword(2) Bayesian estimation
Keyword(3) generalization error
Keyword(4) algebraic geometry
Keyword(5) singularities
1st Author's Name Kazuho WATANABE
1st Author's Affiliation Department of Computer Science, Tokyo Institute of Technology()
2nd Author's Name Sumlo WATANABE
2nd Author's Affiliation P&I Lab., Tokyo Institute of Technology
Date 2002/3/11
Paper # NC2001-149
Volume (vol) vol.101
Number (no) 735
Page pp.pp.-
#Pages 8
Date of Issue