Penalized Partial Least Squares with Applications to B-Spline Transformations and Functional Data
نویسندگان
چکیده
We propose a novel framework that combines penalization techniques with Partial Least Squares (PLS). We focus on two important applications. (1) We combine PLS with a roughness penalty to estimate high-dimensional regression problems with functional predictors and scalar response. (2) Starting with an additive model, we expand each variable in terms of a generous number of B-Spline basis functions. To prevent overfitting, we estimate the model by applying a penalized version of PLS. We gain additional model flexibility by incorporating a sparsity penalty. Both applications can be formulated in terms of a unified algorithm called Penalized Partial Least Squares, which can be computed virtually as fast as PLS using the kernel trick. Furthermore, we prove a close connection of penalized PLS to preconditioned linear systems. In experiments, we show the benefits of our method to noisy functional data and to sparse nonlinear regression models.
منابع مشابه
Spline Estimator for the Functional Linear Regression with Functional Response
The article is devoted to a regression setting where both, the response and the predictor, are random functions defined on some compact sets of R. We consider functional linear (auto)regression and we face the estimation of a bivariate functional parameter. Conditions for existence and uniqueness of the parameter are given and an estimator based on a B-splines expansion is proposed using the pe...
متن کاملHermite Scattered Data Fitting by the Penalized Least Squares Method
Given a set of scattered data with derivative values. If the data is noisy or there is an extremely large number of data, we use an extension of the penalized least squares method of von Golitschek and Schumaker [Serdica, 18 (2002), pp.1001-1020] to fit the data. We show that the extension of the penalized least squares method produces a unique spline to fit the data. Also we give the error bou...
متن کاملPenalized PCA approaches for B-spline expansions of smooth functional data
Functional principal component analysis (FPCA) is a dimension reduction technique that explains the dependence structure of a functional data set in terms of uncorrelated variables. In many applications the data are a set of smooth functions observed with error. In these cases the principal components are difficult to interpret because the estimated weight functions have a lot of variability an...
متن کاملRegularized Least-Squares SPECT Image Reconstruction Using Multiresolution Spatial B-Splines and a Negativity Penalty
We investigated the benefit of incorporating a negativity penalty into a least-squares criterion used to reconstruct 3-D radiotracer distributions in cardiac SPECT studies. B-spline spatial basis functions were used to provide a continuous model for the 3-D tracer distribution. Spline coefficients that tended to have negative values were identified and were constrained to stay near zero with us...
متن کامل