Abstract

In the present paper an axiomatic characterization of non-additive measures of ‘useful’ information associated with a pair of probability distributions of a sample space having utility distribution corresponding to the same number of elements in both probability distributions has been studied. The quantity so obtained under additional suitable postulates leads to the generalized measures of ‘useful’ relative
information, information improvement and J-divergence. Particular cases and
important properties of the measures so obtained have also been studied.

Author: D. S. Hooda and Rakesh Bajaj

Received on: August, 2004

Accepted on: January, 2006