In this paper we study the statistical properties of Laplacian smoothing, a graph-based approach to nonparametric regression. Under standard regularity conditions, we establish upper bounds on the error of the Laplacian smoothing estimator , and a goodness-of-fit test also based on . These upper bounds match the minimax optimal estimation and testing rates of convergence over the first-order Sobolev class , for and ; in the estimation problem, for , they are optimal modulo a factor. Additionally, we prove that Laplacian smoothing is manifold-adaptive: if is an -dimensional manifold with , then the error rate of Laplacian smoothing (in either estimation or testing) depends only on , in the same way it would if were a full-dimensional set in .
View on arXiv