On the probabilistic rationale of I-divergence and J-divergence minimization

TL;DR

This paper provides a probabilistic foundation for I-divergence and J-divergence minimization methods, linking them to likelihood maximization and entropy principles in a unified framework.

Contribution

It offers a new probabilistic rationale for divergence minimization techniques, connecting them to fundamental likelihood and entropy maximization concepts.

Findings

Establishes a probabilistic basis for I-divergence minimization.

Connects J-divergence minimization to Jeffres' entropy maximization.

Unifies divergence minimization with likelihood and entropy principles.

Abstract

A probabilistic rationale for I-divergence minimization (relative entropy maximization), non-parametric likelihood maximization and J-divergence minimization (Jeffres' entropy maximization) criteria is provided.