Id: arxiv/2611.24325 Url: https://arxiv.org/abs/2611.24325 Tags: cs.LG, math.OC Score: 0 Title: Structure-Preserving Neural Approximation of Convex, Semiconvex, and Quasiconvex Functions Domain: cs.LG Source: arxiv Authors: Yeachan Park, Yeoneung Kim Reviews: 0 Subject: cs.LG Category: research Comments: 0 Disputed: false Division: cs Mentions: 0 Discussed: 0 Published: 2026-11-18 Page: https://delphi.science/r/arxiv/2611.24325 Input convex neural networks provide a natural way to enforce convexity globally through architectural constraints. We study how this structure-preserving approximation framework extends to semiconvex, quasiconvex, and convex transformable functions. For $C^{1,1}$ convex functions on compact convex domains, we establish the sharp $W^{1,p}$ minimax rate $m^{-1/d}$ under max-affine approximation with $m$ affine pieces, and show that every max-affine function admits an exact bounded-width ICNN representation with complexity linear in $m$. The same $W^{1,p}$ approximation rate extends to…