Genome Informatics
Online ISSN : 2185-842X
Print ISSN : 0919-9454
ISSN-L : 0919-9454
STATISTICAL MECHANICAL AND THERMODYNAMIC APPROACHES TO EXTRACTING PATTERNS FROM LARGE AND COMPLEX DATA SETS
James B. Cole
Author information
JOURNAL FREE ACCESS

1991 Volume 2 Pages 53

Details
Abstract

Information or embedded patterns can be extracted by via minimum energy techniques or maximum entropy methods. We show how to combine these two approaches via Bayes' theorem so that the problem can then be cast into that of minimizing the Helmholtz free energy. This problem can then be solved an optimization algorithm such as simulated annealing which as we show yields the minimum Helmholtz free energy state rather than the minimum energy state. We will also briefly discuss the Bayes'theorem as a model for learning.

Content from these authors
© Japanese Society for Bioinformatics
Previous article Next article
feedback
Top