Paperlane

no. 12 · Tuesday 10 November 2026

Accelerated primal-dual dynamical systems for convex optimization with nonlinear inequality constraints

Xin He

math.OC · v3

Abstract

In this paper, we propose accelerated inertial primal-dual dynamics for convex optimization problems with nonlinear convex inequality constraints. The proposed dynamics combine Nesterov accelerated vanishing damping $α/t$, a projected augmented-Lagrangian multiplier, and tangent extrapolation of the constraint mapping. Without assuming strong convexity, we establish global well-posedness and simultaneous $\mathcal O(t^{-2})$ decay rates for the constraint violation and the objective residual when $α\geq3$ and the extrapolation parameter satisfies a suitable compatibility condition. Under stricter parameter conditions, with $α>3$, we further prove that the primal-dual trajectory converges to a Karush–Kuhn–Tucker pair, both residuals decay as $o(t^{-2})$. Numerical experiments illustrate the theoretical results.

fields
Optimization and Control, Numerical Analysis
pdf ↗ page arXiv:2609.01415 ↗