A0316
Title: When shape constraints meet kernel machines
Authors: Pierre-Cyril Aubin-Frankowski - INRIA (France)
Zoltan Szabo - LSE (United Kingdom) [presenting]
Abstract: Shape constraints enable one to incorporate prior knowledge into predictive models in a principled way with various successful applications. Including this side information in a hard fashion (e.g, at every point of an interval) for rich function classes, however, is a quite challenging task. We will present a convex optimization framework to encode hard affine constraints on function values and derivatives in the flexible family of kernel machines. The efficiency of the approach is illustrated in joint quantile regression (analysis of aircraft departures).
