PRL: A game theoretic large margin method for interpretable feature learning

Mirko Polato, Guglielmo Faggioli, Fabio Aiolli

Abstract: Highlights•We propose PRL, a large-margin method inspired by preference-learning and game theory.•We propose a parallelized version of the Fictitious Play algorithm for solving game matrices.•We introduce a new feature generation scheme called decision path rules generator.•We present a dynamic budget version of PRL.•PRL generates interpretable hypothesis for solving ranking and classification tasks.