Fitting differential equations to functional data: Principal differential analysis

Abstract

… model variation across observations when N >1. In this sense, it also has the flavor of principal components … In the event that there is no theoretical model at hand, we may use ωj = 0, so …