Por Alexandre Pinto (Professor Auxiliar Convidado, DI/FCUL).

Abstract: Cognitive Modelling encompasses the production of computational models that simulate a variety of human cognitive tasks. Knowledge can be acquired, and represented in a variety of ways including symbolic (e.g., logic), geometric (e.g., feature/value), and connectionist (e.g., neural nets); with correspondingly different kinds of reasoning mechanisms. Current Deep Learning techniques are limited by having fixed topologies, and intermediate components that have no easily understandable global meaning. We present the current version of the Regulated Activation Networks cognitive model aimed at unifying different Knowledge Acquisition, Representation and Reasoning approaches.

Short Bio: Alexandre Miguel Pinto focus his research interests on Cognitive Modelling, namely on Knowledge Acquisition, Representation, Integration, and Reasoning. During his MSc and PhD at FCT-UNL he focused on Knowledge Representation and Reasoning with Non-monotonic Logic and Description Logics. While holding an Invited Assistant Professor position at DEI-FCT-University of Coimbra, from Sep. 2011 until Aug. 2016, he focused on Deep Learning techniques for Knowledge Acquisition. In recent years he has been developing a new Cognitive Model, the Regulated Activation Networks, with applications ranging from pattern recognition to conceptual blending for computational creativity.