INTERACTION INFORMATION IN MULTIVARIATE PROBABILITY DISTRIBUTIONS.
Abstract
It is shown in this note that the entropy of a multivariate distribution can be expressed in terms of the sum of onedimensional marginal entropies, the sum of transmitted information between each pair of component variables, the sum of interaction information in trivariate component distributions, and so on. Using this result, the author gives a class of multivariate distributions having specified component densities and some preassigned association measure between some component variables. Proofs of equations and statements which are not so evident are also given.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 13, 1965
- Accession Number
- AD0623954
Entities
People
- Minoru Sakaguchi
Organizations
- Stanford University