Statistical analysis of regularization constant from bayes, MDL and NIC Points of view

Shun Ichi Amari, Noboru Murata

研究成果: Conference contribution

3 引用 (Scopus)

抜粋

In order to avoid overfitting in neural learning, a regularization term is added to the loss function to be minimized. It is naturMly derived from the Bayesian standpoint. The present paper studies how to determine the regularization constant from the points of view of the empirical Bayes approach, the maximum description length (MDL) approach, and the network information criterion (NIC) approach. The asymptotic statistical analysis is given to elucidate their differences. These approaches are tightly connected with the method of model selection. The superiority of the NIC is shown from this analysis.

元の言語English
ホスト出版物のタイトルBiological and Artificial Computation
ホスト出版物のサブタイトルFrom Neuroscience to Technology - International Work-Conference on Artificial and Natural Neural Networks, IWANN 1997, Proceedings
出版者Springer Verlag
ページ284-293
ページ数10
ISBN(印刷物)3540630473, 9783540630470
出版物ステータスPublished - 1997 1 1
外部発表Yes
イベント4th International Work-Conference on Artificial and Natural Neural Networks, IWANN 1997 - Lanzarote, Canary Islands, Spain
継続期間: 1997 6 41997 6 6

出版物シリーズ

名前Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
1240 LNCS
ISSN(印刷物)0302-9743
ISSN(電子版)1611-3349

Conference

Conference4th International Work-Conference on Artificial and Natural Neural Networks, IWANN 1997
Spain
Lanzarote, Canary Islands
期間97/6/497/6/6

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

フィンガープリント Statistical analysis of regularization constant from bayes, MDL and NIC Points of view' の研究トピックを掘り下げます。これらはともに一意のフィンガープリントを構成します。

  • これを引用

    Amari, S. I., & Murata, N. (1997). Statistical analysis of regularization constant from bayes, MDL and NIC Points of view. : Biological and Artificial Computation: From Neuroscience to Technology - International Work-Conference on Artificial and Natural Neural Networks, IWANN 1997, Proceedings (pp. 284-293). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); 巻数 1240 LNCS). Springer Verlag.