English

Entropy Concentration and the Empirical Coding Game

Information Theory 2008-09-17 v1 Machine Learning math.IT Statistics Theory Methodology Statistics Theory

Abstract

We give a characterization of Maximum Entropy/Minimum Relative Entropy inference by providing two `strong entropy concentration' theorems. These theorems unify and generalize Jaynes' `concentration phenomenon' and Van Campenhout and Cover's `conditional limit theorem'. The theorems characterize exactly in what sense a prior distribution Q conditioned on a given constraint, and the distribution P, minimizing the relative entropy D(P ||Q) over all distributions satisfying the constraint, are `close' to each other. We then apply our theorems to establish the relationship between entropy concentration and a game-theoretic characterization of Maximum Entropy Inference due to Topsoe and others.

Keywords

Cite

@article{arxiv.0809.1017,
  title  = {Entropy Concentration and the Empirical Coding Game},
  author = {Peter Grunwald},
  journal= {arXiv preprint arXiv:0809.1017},
  year   = {2008}
}

Comments

A somewhat modified version of this paper was published in Statistica Neerlandica 62(3), pages 374-392, 2008

R2 v1 2026-06-21T11:17:18.204Z