Stochastic Fixed Points and Nonlinear Perron-Frobenius Theorem

Abstract

We provide conditions for the existence of measurable solutions to the equation $\xi(T\omega)=f(\omega,\xi(\omega))$, where $T:\Omega \rightarrow\Omega$ is an automorphism of the probability space $\Omega$ and $f(\omega,\cdot)$ is a strictly non-expansive mapping. We use results of this kind to establish a stochastic nonlinear analogue of the Perron-Frobenius theorem on eigenvalues and eigenvectors of a positive matrix. We consider a random mapping $D(\omega)$ of a random closed cone $K(\omega)$ in a finite-dimensional linear space into the cone $K(T\omega)$. Under assumptions of monotonicity and homogeneity of $D(\omega)$, we prove the existence of scalar and vector measurable functions $\alpha(\omega)>0$ and $x(\omega)\in K(\omega)$ satisfying the equation $\alpha(\omega)x(T\omega)=D(\omega )x(\omega)$ almost surely.