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

Jun Ichi Inoue, Kazuyuki Tanaka

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

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.

Original languageEnglish
Title of host publicationSlow Dynamics in Complex Systems
Subtitle of host publication3rd International Symposium on Slow Dynamics in Complex Systems
EditorsMichio Tokuyama, Irwin Oppenheim
PublisherAmerican Institute of Physics Inc.
Pages731-734
Number of pages4
ISBN (Electronic)0735401837
DOIs
Publication statusPublished - 2004 Apr 30
Event3rd International Symposium on Slow Dynamics in Complex Systems - Sendai, Japan
Duration: 2003 Nov 32003 Nov 8

Publication series

NameAIP Conference Proceedings
Volume708
ISSN (Print)0094-243X
ISSN (Electronic)1551-7616

Conference

Conference3rd International Symposium on Slow Dynamics in Complex Systems
Country/TerritoryJapan
CitySendai
Period03/11/303/11/8

Fingerprint

Dive into the research topics of 'Dynamics and its stability of Boltzmann-machine learning algorithm for gray scale image restoration'. Together they form a unique fingerprint.

Cite this