汎化誤差最小化による正則化項の設計  [in Japanese] Designing Regularizers by Minimizing Generalization Errors  [in Japanese]

Abstract

汎化能力向上のため, 正則化項を用いた学習が用いられる.汎化能力は正則化パラメータに依存するのみならず, 正則化項にも大きく依存する.本稿では汎化誤差を理論的に導出し, これを用いていずれの正則化項が高い汎化能力を持つかを明らかにし, さらに汎化能力の高い新たな正則化項を見つけたい.もちろん定式化できるのは単純なケースのみであるが, 各正則化項の利害得失はある程度明らかになると考えている.

To improve generalization ability, a regularizer is frequently used. A novel approach proposed here is to regard the estimate of model parameters as a function of those without a regularizer. By minimizing the calcurated generalization error, the optimal function parameters and model parameters can be obtained. In this paper linear regression is adopted to carry out theoretical computation of generalization errors. It also contributes to the design of a new regularizer.

Journal

IEICE technical report. Neurocomputing   [List of Volumes]

IEICE technical report. Neurocomputing 97(624), 55-62, 1998-03-20  [Table of Contents]

The Institute of Electronics, Information and Communication Engineers

References:  4

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.

Preview

Preview

Codes

  • NII Article ID (NAID) :
    110003232805
  • NII NACSIS-CAT ID (NCID) :
    AN10091178
  • Text Lang :
    JPN
  • Article Type :
    ART
  • Databases :
    CJP  NII-ELS 

Share