Presentation 1994/12/16
Learning in a system with a degenerate Fisher information matrix
Sumio Wanana, Kenji Fukumizu,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We study a learning system with a degenerate Fisher information matrix.The optimal parameter that maximizes the logarithm likelihood with an added term goes to the true parameter that maximizes the added term with probability one.The set of true parameters of a layered neural network consists of several manifolds.The asymptotic property of the optimal parameter,that of the training loss,and the prediction loss are also analyzed.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Fisher Information Matrix / Consistency / Layered Neural Network
Paper # NC94-56
Date of Issue

Conference Information
Committee NC
Conference Date 1994/12/16(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) Learning in a system with a degenerate Fisher information matrix
Sub Title (in English)
Keyword(1) Fisher Information Matrix
Keyword(2) Consistency
Keyword(3) Layered Neural Network
1st Author's Name Sumio Wanana
1st Author's Affiliation Ricoh Co.,Ltd.()
2nd Author's Name Kenji Fukumizu
2nd Author's Affiliation Ricoh Co.,Ltd.
Date 1994/12/16
Paper # NC94-56
Volume (vol) vol.94
Number (no) 417
Page pp.pp.-
#Pages 8
Date of Issue