PRIOR PROBABILITY, TWO-PERSON GAMES AND INFORMATION THEORY,
Abstract
The prior probability of a certain variable B is determined using all the available deterministic knowledge such as (1) the variable B may not be observable, but only a related variable Z may be observable so that P(Z/B) is known; (2) the probability distributions that can be assumed by B are limited by the inequalities E(h sub k (B)) > or = o where h sub k (B) are known functions. The prior probability is determined by a two-person zero-sum game which deals with the gambling on the outcomes of the observable variable Z. The prior probability distribution determined in this manner is the same as that obtained by maximizing the average uncertainty associated with the variable B given that variable Z is observable. An expression will be obtained for the uncertainty function starting with a set of axioms and the expression is the same as the Shannon mutual information I(B;Z) between the variables B and Z. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1969
- Accession Number
- AD0698773
Entities
People
- R. L. Kashyap
Organizations
- Purdue University