Presentation 1997/11/21
Method for finding the Optimal Number of Dimension of Subspace Using Akaike Information Criteria
Hitoshi Sakano, Naoki Mukawa,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Subspace method has been considered important method for pattern recognition. However, how to determine the number of dimension of subspace is still an unsolved problem. In most applications, it is determined experientially and less theoretical analysis is reported. In this paper, we propose a method that can find the optimal dimensionally of subspace through using Akaike Information Criteria (AIC). The effectiveness of proposed method is examined with two experiments of character recognition and face verification. Experimental results show that our proposal is more effective than faithfullness method to find a suitable dimension for a subspace.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) AIC / Subspace Method / Character Recognition / Face Verification
Paper # PRMU97-173
Date of Issue

Conference Information
Committee PRMU
Conference Date 1997/11/21(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 Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Method for finding the Optimal Number of Dimension of Subspace Using Akaike Information Criteria
Sub Title (in English)
Keyword(1) AIC
Keyword(2) Subspace Method
Keyword(3) Character Recognition
Keyword(4) Face Verification
1st Author's Name Hitoshi Sakano
1st Author's Affiliation Laboratry for Information Technology NTT DATA CORPORATION()
2nd Author's Name Naoki Mukawa
2nd Author's Affiliation Laboratry for Information Technology NTT DATA CORPORATION
Date 1997/11/21
Paper # PRMU97-173
Volume (vol) vol.97
Number (no) 387
Page pp.pp.-
#Pages 8
Date of Issue