Statistical Mechanics of Time-Domain Ensemble Learning

• Miyoshi Seiji

  Department of Electronic Engineering, Kobe City College of Technology

• Uezu Tatsuya

  Graduate School of Humanities and Sciences, Nara Women’s University

• Okada Masato

  Division of Transdisciplinary Sciences, Graduate School of Frontier Sciences, The University of Tokyo RIKEN Brain Science Institute JST PRESTO

Abstract

Conventional ensemble learning combines students in the space domain. On the other hand, in this paper, we combine students in the time domain and call it time-domain ensemble learning. We analyze the generalization performance of time-domain ensemble learning in the framework of on-line learning using a statistical mechanical method. We use a model in which both the teacher and the student are linear perceptrons with noises. Time-domain ensemble learning is twice as effective as conventional space-domain ensemble learning.

Journal

• Journal of the Physical Society of Japan

  Journal of the Physical Society of Japan 75 (8), 084007-, 2006

  THE PHYSICAL SOCIETY OF JAPAN

Keywords

• ensemble learning
• on-line learning
• generalization error
• statistical mechanics

Details

• CRID: 1390282679166167680
• NII Article ID: 110004799205
• NII Article ID: 210000106275
• NII Book ID: AA00704814
• DOI: 10.1143/jpsj.75.084007
• BIBCODE: 2006JPSJ...75h4007M
• ISSN: 13474073
• ISSN: 00319015
• NDL BIB ID: 8020023
• Web Site: http://id.ndl.go.jp/bib/8020023
• Web Site: https://ndlsearch.ndl.go.jp/books/R000000004-I8020023
• Web Site: https://journals.jps.jp/doi/pdf/10.1143/JPSJ.75.084007
• Text Lang: en
• Data Source:
  • JaLC
  • NDL
  • Crossref
  • CiNii Articles
  • KAKEN
• Abstract License Flag: Disallowed