The aim of this project is to develop parsimonious models for regression and classification based on kernel methods, to provide enhanced visualisation and generalisation in empirical modelling.