{"id":"arxiv/2606.28641","url":"https://arxiv.org/abs/2606.28641","tags":["math.ST"],"score":0,"title":"Revisiting local regression: shape regularity, uniform rates, and the limits of random splits","domain":"math.ST","source":"arxiv","authors":["Jérémy Bettinger","François Portier","Adrien Saumard"],"reviews":0,"subject":"math.ST","summary":"Considering pointwise and sup-norm estimation, we analyze the non-asymptotic behavior of local averaging estimators for Lipschitz regression functions. Building on a general deviation bound for estimators based on a VC family of localizing sets, we introduce the notion of shape-regular local maps, where averaging is performed over sets with an almost isotropic geometry. Our main message is a characterization: shape regularity is both necessary and sufficient to attain optimal rates, up to logarithmic factors. Necessity is established non-asymptotically through an explicit anisotropic…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-06-26","reactions":{}}