In this paper we build upon the Multiple Kernel Learning (MKL) framework. We rewrite the problem in the standard MKL formulation which leads to a Semi-Infinite Program. We devise a new algorithm to
