Id: arxiv/2602.04479 Url: https://arxiv.org/abs/2602.04479 Tags: math.OC Score: 0 Title: Decentralized Optimization with Mixed Affine Constraints Domain: math.OC Source: arxiv Authors: Demyan Yarmoshik, Nhat Trung Nguyen, Alexander Rogozin, Alexander Gasnikov Reviews: 0 Subject: math.OC Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-02-04 Page: https://delphi.science/r/arxiv/2602.04479 This paper considers decentralized optimization of convex functions with mixed affine equality constraints involving both local and global variables. Constraints on global variables may vary across different nodes in the network, while local variables are subject to coupled and node-specific constraints. Such problem formulations arise in machine learning applications, including federated learning and multi-task learning, as well as in resource allocation and distributed control. We analyze this problem under smooth and non-smooth assumptions, considering both strongly convex and general…