Fast Compact Algorithms and Software for Spline Smoothing

By Howard L. Weinert

Subjects: Mathematical Software, Statistics and Computing/Statistics Programs, Engineering mathematics, Appl.Mathematics/Computational Methods of Engineering, Computer science, Computational Science and Engineering, Computer software, Mathematical statistics, Statistics, Image and Speech Processing Signal

Description: <i>Fast Compact Algorithms and Software for Spline Smoothing</i> investigates algorithmic alternatives for computing cubic smoothing splines when the amount of smoothing is determined automatically by minimizing the generalized cross-validation score. These algorithms are based on Cholesky factorization, QR factorization, or the fast Fourier transform. All algorithms are implemented in MATLAB and are compared based on speed, memory use, and accuracy. An overall best algorithm is identified, which allows very large data sets to be processed quickly on a personal computer.