DCCMAPI/CCET

Uma banca de PROPOSTA DE TESE/DISSERTAÇÃO de DOUTORADO foi cadastrada pelo programa.
DISCENTE: MATHEUS CHAVES MENEZES
DATA: 25/03/2022
HORA: 14:00
LOCAL: Google meeting
TÍTULO: A Computational Neuro-inspired Framework to Study Animal Behavior Through Spatial Navigation
PALAVRAS-CHAVES: Latent learning, Cognitive maps, Predictive maps, Successor representation, RatSLAM, Animals’ behavior.
PÁGINAS: 64
GRANDE ÁREA: Ciências Exatas e da Terra
ÁREA: Ciência da Computação
SUBÁREA: Metodologia e Técnicas da Computação
ESPECIALIDADE: Banco de Dados
RESUMO: The latent learning sets that animals improve their spatial navigation tasks if they are previously exposed to the environment. These observations made the theory that animals hold a cognitive map of the environment. The findings supported this theory in the neuroscience of hippocampus and entorhinal cortex cells that selectively fire on space. However, recent studies brought new evidence that these cognitive maps could work in a predictive manner. The Successor Representation algorithm can explain this predictive representation. Many studies on the predictive maps rely on a given environment model with all modeled states. In this study, we propose a new way to acquire the states of the environment by a biologically plausible mapping algorithm called RatSLAM. We employed the RatSLAM in a computational framework to study animals’ behavior. Besides that, we solve real-world problems addressed to RatSLAM. As preliminary results, our framework could replicate the latent learning experiment. With this, we might contribute with a framework to study animals’ flexible behavior in a neurologically plausible way.
MEMBROS DA BANCA:
Presidente - 1091306 - ALEXANDRE CESAR MUNIZ DE OLIVEIRA
Interno - 2044484 - AREOLINO DE ALMEIDA NETO
Interno - 1564359 - LUCIANO REIS COUTINHO
Externo à Instituição - DANTE AUGUSTO COUTO BARONE - UFRGS
Externo à Instituição - EDISON PIGNATON DE FREITAS - UFRGS
Co-orientador - 000.000.000-00 - SEN CHENG