{"id":"arxiv/2610.16429","url":"https://arxiv.org/abs/2610.16429","tags":["math.OC","cs.LG"],"score":0,"title":"Taming Nonconvexity with Epigraph Lifting","domain":"math.OC","source":"arxiv","authors":["Clara Lage","Hok Shing Wong","Matthias Ehrhardt"],"reviews":0,"subject":"math.OC","summary":"We consider inverse problems regularized by a trained neural network with \\emph{convex activations} and \\emph{unrestricted weights}. The resulting optimization problem is generally nonconvex, and gradient-based methods may converge to poor local minima. We introduce a \\emph{penalized epigraph lifting} of the network and show that the lifted problem admits a \\emph{difference-of-convex (DC) decomposition}. For Lipschitz continuous activations, we establish \\emph{finite exactness}: the original objective is recovered exactly at a finite and explicitly computable penalty threshold $\\barρ$.…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-12","reactions":{}}