{"id":"arxiv/2601.10357","url":"https://arxiv.org/abs/2601.10357","tags":["stat.ME","math.ST"],"score":0,"title":"Model-Agnostic and Uncertainty-Aware Dimensionality Reduction in Supervised Learning","domain":"stat.ME","source":"arxiv","authors":["Yue Yu","Guanghui Wang","Liu Liu","Changliang Zou"],"reviews":0,"subject":"stat.ME","summary":"Dimension reduction is a fundamental tool for analyzing high-dimensional data in supervised learning. Traditional methods for estimating intrinsic order often prioritize model-specific structural assumptions over predictive utility. This paper introduces predictive order determination (POD), a model-agnostic framework that determines the minimal predictively sufficient dimension by directly evaluating out-of-sample predictiveness. POD quantifies uncertainty via error bounds for over- and underestimation and achieves consistency under mild conditions. By unifying dimension reduction with…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-01-15","reactions":{}}