{"id":"arxiv/2606.17285","url":"https://arxiv.org/abs/2606.17285","tags":["math.OC","cs.DS"],"score":0,"title":"An Adaptive Proximal Framework for Stochastic Weakly Convex Optimization","domain":"math.OC","source":"arxiv","authors":["Miaolan Xie"],"reviews":0,"subject":"math.OC","summary":"Many nonsmooth, nonconvex objectives in learning and signal recovery are $ρ$-weakly convex. We minimize such a function when the weak-convexity parameter $ρ$ is unknown and only stochastic candidate and function-difference information is available. We propose the Adaptive Prox-Guided Scheme (APS), a single-trial framework that adapts the proximal parameter online and bidirectionally through a noisy descent test. APS attains a high-probability $O(\\varepsilon^{-2})$ iteration bound for Moreau-envelope stationarity. The result allows biased, heavy-tailed function-difference estimates, while the…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-06-15","reactions":{}}