Aspectos de mapas caóticos acoplados para processamento de informações / Aspects of chaotic coupled maps for information processing

AUTOR(ES)
DATA DE PUBLICAÇÃO

2002

RESUMO

A globally coupled map (GCM) model is a network of interconnected chaotic elements. In this work we investigate models based in the GCM and in the Hopfield network. Through modifications, like changing the local dynamics of the GCM (S-GCM) and adding self-feedback to the processing element of the Hopfield network, it s possible to use the models as an assocíative memory. Also through the exploration of the chaotic dynamics in such a way to prevent spurious mínimums, ít is possible to propose two new networks that firstly were applied in optimization problems, that are also able to process information in a associative manner. One of them is a modification of the S-GCM with a determined self-feedback rate and the other, is based in a specific GCM s feature in respect to the changing of the network s attractor through pertubatíons. Finally, the results of the networks are numerically compared based on the size of the attraction basin, restrictions imposed by the arquítecture, memory capacity, association time, and mínímization capacity.

ASSUNTO(S)

inteligência artificial. key words: dynamical systems sistemas dinâmicos information processing atratores estranhos chaos neural nets optimization artificial inteligence processamento de informações caos strange attractors otimização redes neurais

ACESSO AO ARTIGO

Documentos Relacionados