Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
1866586 | Physics Letters A | 2006 | 6 Pages |
Abstract
This Letter considers a geometrical structure on the parameter space of the belief propagation algorithm on Bayesian networks. The statistical manifold of posterior distributions is introduced, and the expression for the information metric on the manifold is derived. The expression is used to construct a cost function which can be regarded as a measure of the distance in the parameter space.
Related Topics
Physical Sciences and Engineering
Physics and Astronomy
Physics and Astronomy (General)
Authors
Yodai Watanabe,