{"id":"arxiv/2609.10732","url":"https://arxiv.org/abs/2609.10732","tags":["math.OC"],"score":0,"title":"Operator Splitting Methods with Online Scaling","domain":"math.OC","source":"arxiv","authors":["Wanyu Zhang","Wenzhi Gao","Madeleine Udell"],"reviews":0,"subject":"math.OC","summary":"This paper develops a principled framework for automatically tuning preconditioners in operator splitting methods, including forward-backward splitting, Douglas-Rachford splitting, and the alternating direction method of multipliers. We interpret these splitting methods as preconditioned gradient descent applied to their corresponding envelope functions and use online learning to adaptively learn the preconditioner during optimization. Our framework achieves asymptotically faster convergence rates and accelerates the solution of important optimization problems, including Lasso,…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-09","reactions":{}}