Asymptotic Behavior of the Expectation Value of Permanent Products, a Sequel

TL;DR

This paper investigates the asymptotic behavior of the expectation value of products of permanents in random matrix ensembles, comparing approximations and establishing limits for Bernoulli matrices as dimensions grow large.

Contribution

It extends previous work by analyzing the expectation of permanent products in Bernoulli ensembles and compares these with earlier approximations, providing asymptotic equivalences and conjectures.

Findings

Asymptotic equivalence of expectation values in different ensembles

Limit formulas for products of permanents as matrix size grows

Conjecture on the accuracy of approximations in large r limit

Abstract

Continuing the computations of the previous paper,[1], we calculate another approximation to the expectation value of the product of two permanents in the ensemble of 0-1 n x n matrices with like row and column sums equal r uniformly weighted. Here we consider the Bernoulli random matrix ensemble where each entry independently has a probability p=r/n of being one, otherwise zero. We denote the expectations of the approximation ensemble of [1] by E, and the expectations of the present approximation ensemble, the Bernoulli random matrix ensemble, by E*. One has for these lim_{r to infinity}( lim_{n to infinity} (1/n) ln(E(perm_m(A))) -lim_{n to infinity} (1/n) ln(E*(perm_m(A))) ) = 0 and lim_{n to infinity} (1/n) ln(E(perm_m(A)perm_m'(A))) = lim_{n to infinity} (1/n) ln(E(perm_m(A))) + lim_{n to infinity} (1/n) ln(E(perm_m'(A))) Here and in all such formulas the subscripts m,m' are assumed proportional to n. It seems likely to us that lim_{r to infinity}( lim_{n to infinity} (1/n) ln(E*(perm_m(A)perm_m'(A))) - lim_{n to infinity} (1/n) ln(E*(perm_m(A))) + - lim_{n to infinity} (1/n) ln(E*(perm_m'(A))) ) = 0 We believe: "E gives us the `correct' expectations in these equations, and E* is only `correct' in the r to infinity limit."