Discretized Laplacian Smoothing by Fourier Methods

From MaRDI portal





A class of multidimensional smoothers is defined as minimizers of the penalized likelihoods of a certain form [see \textit{D. D. Cox}, SIAM J. Numer. Anal. 21, 789-813 (1984; Zbl 0581.65012) and \textit{G. Wahba}, Spline models in statistics, CBMS Regional Conference Series in Applied Mathematics, Philadelphia: SIAM (1990)]. Multidimensional smoothing procedures based on the foregoing class encounter major computational difficulties for dimension greater than 1. A discretized version of the multidimensional smoother with Laplacian penalty is introduced. The author shows that a well-chosen discretization can lead to considerable simplifications in computations without appreciably compromising performance characteristics. The paper only considers the case where the region of interest is a \(d\)-dimensional rectangle.











This page was built for publication: Discretized Laplacian Smoothing by Fourier Methods

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3984601)