{"id":"arxiv/2610.24293","url":"https://arxiv.org/abs/2610.24293","tags":["cs.LG","math.FA","stat.ML"],"score":0,"title":"Constructive Universal Approximation by Infinite-Dimensional Neural Networks via Gaussian-Weighted Ridgelet Transform","domain":"cs.LG","source":"arxiv","authors":["Sho Sonoda","Luca Galimberti","Yuanchao Xu","Isao Ishikawa","Masahiro Ikeda"],"reviews":0,"subject":"cs.LG","summary":"Neural networks whose inputs are functions, as in operator learning, are usually shown to be universal by existence arguments, either by reduction to finite-dimensional networks or by Stone--Weierstrass-type density; these prove that approximants exist but say little about how the parameters depend on the target. We take a constructive route. We extend the ridgelet transform, an integral transform related to the Fourier, Radon, and wavelet transforms, to continuous maps on a separable infinite-dimensional Hilbert space, replacing the Lebesgue measure that is unavailable in infinite…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-10-16","reactions":{}}