Information Theory

Ordering of Information Criteria for Discrimination of Probability Distributions

A. A. Amosov, V. V. Kolpakov


Abstract: The authors introduce an axiom system that characterizes so-called divergences, i.e., a class of information measures for discriminating probability distributions with specified properties. A general analytic form for the functional in question is given. A number of inequalities are established between the most widely used information criteria for discriminating probability distributions, these being linear functions of the divergences.

UDC: 621.391.1:519.21

Received: 21.01.1974
Revised: 14.04.1975

 English version: Problems of Information Transmission, 1976, 12:4, 245–250

Bibliographic databases:

• 
•