Dynamics and its stability of Boltzmann-machine learning algorithm for gray scale image restoration

Jun Ichi Inoue, Kazuyuki Tanaka

研究成果: Conference contribution

抄録

Dynamic behavior and its stability of Boltzmann-machine learning algorithm for Bayesian gray scale image restoration are investigated. We derive the differential equations by which we attempt to maximize the marginal likelihood function with respect to hyper-parameters. Average-case performance and linear stability of the algorithm are evaluated exactly at the mean-field level. We conclude that the solution of the Boltzmann-machine learning equation is asymptotically stable as long as the solution is identical to the correct value of the hyper-parameters.

本文言語English
ホスト出版物のタイトルSlow Dynamics in Complex Systems
ホスト出版物のサブタイトル3rd International Symposium on Slow Dynamics in Complex Systems
編集者Michio Tokuyama, Irwin Oppenheim
出版社American Institute of Physics Inc.
ページ731-734
ページ数4
ISBN(電子版)0735401837
DOI
出版ステータスPublished - 2004 4 30
イベント3rd International Symposium on Slow Dynamics in Complex Systems - Sendai, Japan
継続期間: 2003 11 32003 11 8

出版物シリーズ

名前AIP Conference Proceedings
708
ISSN(印刷版)0094-243X
ISSN(電子版)1551-7616

Other

Other3rd International Symposium on Slow Dynamics in Complex Systems
国/地域Japan
CitySendai
Period03/11/303/11/8

ASJC Scopus subject areas

  • 物理学および天文学(全般)

フィンガープリント

「Dynamics and its stability of Boltzmann-machine learning algorithm for gray scale image restoration」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル