Saddle Flow Dynamics: Observable Certificates and Separable Regularization

TL;DR

This paper introduces an observable certificate for saddle flow dynamics that guarantees convergence to saddle points under minimal conditions and extends to distributed linear programming solutions.

Contribution

It presents a novel observable certificate based on observability theory and a separable regularization method that broadens convergence guarantees for saddle flow dynamics.

Findings

Certificate guarantees asymptotic convergence under minimal convexity-concavity.

Separable regularization ensures convergence with relaxed assumptions.

Applicable to distributed linear programming solutions.

Abstract

This paper proposes a certificate, rooted in observability, for asymptotic convergence of saddle flow dynamics of convex-concave functions to a saddle point. This observable certificate directly bridges the gap between the invariant set and the equilibrium set in a LaSalle argument, and generalizes conventional conditions such as strict convexity-concavity and proximal regularization. We further build upon this certificate to propose a separable regularization method for saddle flow dynamics that makes minimal requirements on convexity-concavity and yet still guarantees asymptotic convergence to a saddle point. Our results generalize to saddle flow dynamics with projections on the vector field and have an immediate application as a distributed solution to linear programs.