The Scaling Mean and a Law of Large Permanents

TL;DR

This paper introduces a new ergodic theorem for permanental means of dynamically generated matrices, linking them to a generalized scaling mean and unifying various classical means.

Contribution

It establishes a Law of Large Permanents connecting permanental means with a generalized scaling mean in an ergodic setting, extending classical results.

Findings

Proves a pointwise ergodic theorem for permanental means.

Defines a generalized scaling mean for functions.

Reobtains a classical limit formula for symmetric means.

Abstract

In this paper we study two types of means of the entries of a nonnegative matrix: the \emph{permanental mean}, which is defined using permanents, and the \emph{scaling mean}, which is defined in terms of an optimization problem. We explore relations between these two means, making use of important results by Ergorychev and Falikman (the van~der~Waerden conjecture), Friedland, Sinkhorn, and others. We also define a scaling mean for functions in a much more general context. Our main result is a Law of Large Permanents, a pointwise ergodic theorem for permanental means of dynamically defined matrices that expresses the limit as a functional scaling mean. The concepts introduced in this paper are general enough so to include as particular cases certain classical types of means, as for example symmetric means and Muirhead means. As a corollary, we reobtain a formula of Hal\'asz and Sz\'ekely for the limit of the symmetric means of a stationary random process.