A0165
Title: Nonparametric regression on contaminated functional predictor with application to hyperspectral data
Authors: Frederic Ferraty - Mathematics Institute of Toulouse (France) [presenting]
Abstract: The focus is on scalar-on-function nonparametric regression when only a contaminated version of the functional predictor is observable at some measurement grid. To override this common setting, a kernel presmoothing step is achieved on the noisy functional predictor. Then, the kernel estimator of the regression operator is built using the smoothed functional covariate instead of the original corrupted ones. The rate of convergence is stated for this nested-kernel estimator (i.e. the smoothing parameter of the presmoothing stage minimizes some predictive criterion) with a special attention on high-dimensional setting (i.e the size of the measurement grid is much larger than the sample size). The proposed method is applied on simulated datasets in order to assess its finite-sample properties. Our methodology is further illustrated on a real hyperspectral dataset involving a supervised classification problem.
