Presentation	1996/2/3 Properties of Overload Learning : Case Study of Learning Identity-function by Linear Networks Itsuki NODA,
PDF Download Page	PDF download Page Link
Abstract(in Japanese)	(See Japanese page)
Abstract(in English)	The relation between parameters of overload learning (OLL) and internal representation acquired by OLL is reported. OLL has proposed for reducing redundant representation in the patterns on hidden layers. It is, however, not clear how to decide values of parameters used in OLL. In this article I describe some results of theoretical analysis of OLL in the case of learning identity-function by liner networks, and derive some properties of OLL. The properties are confirmed trough some experiments. Moreover, I show that OLL with nonlinear units also has similar properties through experiments.
Keyword(in Japanese)	(See Japanese page)
Keyword(in English)	overload learning / internal representation / identity function / feature abstraction
Paper #	NC95-113
Date of Issue

Conference Information
Committee	NC
Conference Date	1996/2/3(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)	Properties of Overload Learning : Case Study of Learning Identity-function by Linear Networks
Sub Title (in English)
Keyword(1)	overload learning
Keyword(2)	internal representation
Keyword(3)	identity function
Keyword(4)	feature abstraction
1st Author's Name	Itsuki NODA
1st Author's Affiliation	Electrotechnical Laboratory()
Date	1996/2/3
Paper #	NC95-113
Volume (vol)	vol.95
Number (no)	506
Page	pp.pp.-
#Pages	8
Date of Issue