Implementação de redes neuronais

• Autor(a) principal: Azevedo, Moacyr Henrique Cruz de
• Data de Publicação: 1991
• Tipo de documento: Relatório
• Idioma: por
• Título da fonte: Repositório Institucional da UFRJ
• Texto Completo: http://hdl.handle.net/11422/1551
• topic: Redes neuronais (Computadores); CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO

The first neural networks have been simulated on sequential computers. As neural networks with some practical importance have hundreds of neurons, the simulation time has become a difficult barrier for the researchers. To avoid this problem, some researchers began to simulate on existing parallel computers. Other researchers chose to develop a specialized hardware, the NEURO-COMPUTERS, that can be eletronic or eletro-optic. The eletronic implementation can use commercial or special chips, the NEURAL-CHIPS, and have digital or analog implementation. The eletro-optic implementation is called OPTICAL NEURAL-COMPUTERS. This work discribes four implementations for neural networks: analog neural-computer, digital neural-computer, optical neural-computer, and simulation on a parallel computer.