Location: A226
Abstract: The seamless integration of defeasible reasoning and probabilistic settings appear as a long-standing pursuit to capture qualitative as well as quantitative uncertainty in the same frame. In this direction, I will present a novel class of Boltzmann called “argumentative Boltzmann machines” where the possible states of a conventional restricted Boltzmann machine are constrained by some prior argumentative knowledge.
About the speaker: Dr. Régis Riveret is Marie-Curie Fellow in the Intelligent Systems and Networks Group at the Department of Electrical and Electronic Engineering in Imperial College. His research interests deal with probabilistic rule-based argumentation and its use to study norm-governed learning agents. Before this position, he worked in the industry (Amadeus) on the management and application of executable specification of travel and travellers policy rules. He holds an Erasmus degree in Physics (University of Bristol – Université de Caen), a Master degree in Telecommunication (Télécom Bretagne) and a Phd in Computer Sciences & Law (Università di Bologna).
