Id: arxiv/2611.20437 Url: https://arxiv.org/abs/2611.20437 Tags: math.OC, cs.LG Score: 0 Title: On the Convergence of Continuous-Time AdamW Domain: math.OC Source: arxiv Authors: Bangjie Wang, Ning Ning Reviews: 0 Subject: math.OC Category: research Comments: 0 Disputed: false Division: math Mentions: 0 Discussed: 0 Published: 2026-11-16 Page: https://delphi.science/r/arxiv/2611.20437 Continuous-time AdamW provides a framework for understanding whether an optimizer widely used to train large-scale neural networks settles to an equilibrium or sustains persistent motion. We study this long-time behavior while retaining bias correction and fixed positive weight decay. Under suitable assumptions, we establish that the dynamics are well defined for all time and remain bounded when the decay is sufficiently small. Our main result connects their limiting behavior to the geometry of the objective: trajectories approach a locally invariant manifold whose dimension equals the…