構造学習における初期値設定に関する検討  [in Japanese] A Study on Initialization for Structural Learning with Forgetting  [in Japanese]

Abstract

本報告では先に提案した初期値設定法が忘却付き構造学習によるニューラルネットの構造決定にどのような影響を及ぼすかを検討する.この初期値設定法は2つのstepから構成されており,中間ユニットに対しては,そのユニットのなす分離超平面が入力集合の中心を通るように設定し,出力ユニットの荷重は全てゼロで設定する。アヤメの分類問題を対象としたシミュレーション実験の結果,提案法を用いることで,ネット構造が改善され,汎化能力も向上することを確認した.

This paper studies how our proposed initialization effects the structure determination of neural networks by structural learning with forgetting. The proposed initialization consists of two steps: weights of hidden units are initialized so that their hyperplanes should pass through the center of input pattern set, and those of output units are initialized to zero. From simulation result performed on Iris problem, it was confirmed that the initialization gives better network structure and higher generalization ability.

Journal

IEICE technical report. Neurocomputing   [List of Volumes]

IEICE technical report. Neurocomputing 99(473), 39-45, 1999-11-26  [Table of Contents]

The Institute of Electronics, Information and Communication Engineers

References:  5

You must have a user ID to see the references.If you already have a user ID, please click "Login" to access the info.New users can click "Sign Up" to register for an user ID.

Cited by:  3

You must have a user ID to see the cited references.If you already have a user ID, please click "Login" to access the info.New users can click "Sign Up" to register for an user ID.

Preview

Preview

Codes

  • NII Article ID (NAID) :
    110003234379
  • NII NACSIS-CAT ID (NCID) :
    AN10091178
  • Text Lang :
    JPN
  • Article Type :
    Journal Article
  • ISSN :
    09135685
  • NDL Article ID :
    4934830
  • NDL Source Classification :
    ZN33(科学技術--電気工学・電気機械工業--電子工学・電気通信)
  • NDL Call No. :
    Z16-940
  • Databases :
    CJP  CJPref  NDL  NII-ELS 

Share