Bannière Faculté des sciences DIC
Titre : CONFÉRENCE: Machine Learning Methods for Word Learning and Perceptual Categorization
Auteur : Hans Bassani
Nombre : 18/17
Etat : Possibilité de réserver
Début : Lundi, 22 Mai, 2017 à 12:00
Lieu : SU 1550 (Amphitheatre)
Réservation : 12


Concept acquisition is a central ability required for many cognitive tasks, including word learning for language acquisition. There is a significant amount of theoretical work suggesting that certain types of concepts can be learned through the categorization of perceptions. The creation of a computational model for perception categorization, capable of dealing with real-world data, could greatly advance research in the fields mentioned above. It could allow us to replicate in silicon, experiments carried out with human beings and evaluating theoretical models and hypotheses about related phenomena. In this talk, we will describe a perception categorization model that was developed based on state-of-the-art machine learning methods. The proposed model is capable of handling high-dimensional real-world inputs such as image and audio and create categories of perceptions that are consistent with simple concrete concepts expected to be acquired by human subjects with the provided input data. The model has been applied to replicate cross-situational word learning experiments carried learning experiments carried out with human subjects in various situations, displaying similar word learning patterns


Hansenclever de França Bassani,

Professor Adjunto, Universidade Federal de Pernambuco, Centro de Informática - CIn, Departamento de Sistemas de Computação