Redes Neurais Atratoras em Topologias Complexas

• Ano de defesa: 2006
• Autor(a) principal: Galhardo, Carlos Eduardo Cardoso
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Programa de Pós-graduação em Física; Física
• Palavras-chave em Português: Rede neural; Topologia; Neurociências; Mecânica estatística; CNPQ::CIENCIAS EXATAS E DA TERRA::FISICA
• Link de acesso: https://app.uff.br/riuff/handle/1/18934

In this work we present some modifications in the Hopfield model for attractor neural networks. First, in order to bring the model closer to a biological network of neurons, we introduce a controlled dilution generating complex topologies. This modification is supported by experimental evidences [1, 2]. Then, study the relation between eficiency and the topologies, in the recovery stored patterns. We study how the network topology play a role in the model, by introducing correlated patterns and studing the performance of each topology as the correlation between patterns is increased.