B0807
Title: Adaptive functional scores
Authors: Sunny Wang - ENSAI (France) [presenting]
Valentin Patilea - CREST-Ensai (France)
Abstract: The infinite-dimensional nature of functional data necessitates that some form of dimensionality reduction is performed, often carried out in practice with principal components analysis (PCA). Random functions are represented using a linear combination of the eigenbasis, where these coefficients are referred to as the principal component scores. Although several methods exist for computing functional scores, methods which automatically adapt to the regularity of the sample paths are relatively underdeveloped. An adaptive method is proposed for computing the scores based on adaptive constructions of the eigen elements. These individual scores adapt to a wide class of functions where the sample paths are not required to be differentiable. Finite sample properties of the methods are explored with a versatile simulator.
