A maximin characterization of the escape rate of nonexpansive mappings in metrically convex spaces
Stephane Gaubert, Guillaume Vigeral

TL;DR
This paper provides a maximin characterization of the escape rate for non-expansive mappings in metric spaces with curvature conditions, generalizing classical spectral radius results and applying to stochastic games.
Contribution
It introduces a novel maximin formula involving horofunctions for the escape rate of non-expansive maps under curvature assumptions, extending classical spectral and fixed point theorems.
Findings
Generalizes Collatz-Wielandt spectral radius characterization
Includes a new theorem for non-expansive maps in Banach spaces
Refines results on order-preserving homogeneous maps
Abstract
We establish a maximin characterisation of the linear escape rate of the orbits of a non-expansive mapping on a complete (hemi-)metric space, under a mild form of Busemann's non-positive curvature condition (we require a distinguished family of geodesics with a common origin to satisfy a convexity inequality). This characterisation, which involves horofunctions, generalises the Collatz-Wielandt characterisation of the spectral radius of a non-negative matrix. It yields as corollaries a theorem of Kohlberg and Neyman (1981), concerning non-expansive maps in Banach spaces, a variant of a Denjoy-Wolff type theorem of Karlsson (2001), together with a refinement of a theorem of Gunawardena and Walsh (2003), concerning order-preserving positively homogeneous self-maps of symmetric cones. An application to zero-sum stochastic games is also given.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
