Last-iterate Convergence of ADMM on Multi-affine Quadratic Equality Constrained Problem

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chao, Yutong, Ciebielski, Michal, Etesami, Jalal, Khadiv, Majid
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911508850540544
author Chao, Yutong
Ciebielski, Michal
Etesami, Jalal
Khadiv, Majid
author_facet Chao, Yutong
Ciebielski, Michal
Etesami, Jalal
Khadiv, Majid
contents In this paper, we study a class of non-convex optimization problems known as multi-affine quadratic equality constrained problems, which appear in various applications--from generating feasible force trajectories in robotic locomotion and manipulation to training neural networks. Although these problems are generally non-convex, they exhibit convexity or related properties when all variables except one are fixed. Under mild assumptions, we prove that the alternating direction method of multipliers (ADMM) converges when applied to this class of problems. Furthermore, when the "degree" of non-convexity in the constraints remains within certain bounds, we show that ADMM achieves a linear convergence rate. We validate our theoretical results through practical examples in robotic locomotion.
format Preprint
id arxiv_https___arxiv_org_abs_2603_11919
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Last-iterate Convergence of ADMM on Multi-affine Quadratic Equality Constrained Problem
Chao, Yutong
Ciebielski, Michal
Etesami, Jalal
Khadiv, Majid
Optimization and Control
90C26
G.1.6
In this paper, we study a class of non-convex optimization problems known as multi-affine quadratic equality constrained problems, which appear in various applications--from generating feasible force trajectories in robotic locomotion and manipulation to training neural networks. Although these problems are generally non-convex, they exhibit convexity or related properties when all variables except one are fixed. Under mild assumptions, we prove that the alternating direction method of multipliers (ADMM) converges when applied to this class of problems. Furthermore, when the "degree" of non-convexity in the constraints remains within certain bounds, we show that ADMM achieves a linear convergence rate. We validate our theoretical results through practical examples in robotic locomotion.
title Last-iterate Convergence of ADMM on Multi-affine Quadratic Equality Constrained Problem
topic Optimization and Control
90C26
G.1.6
url https://arxiv.org/abs/2603.11919