A Reconfigurable Computer REOMP
Alessandro Noriaki Ide, José Hiroki Saito · 2001
This work describes a proposal of reconfigurable computer, and their application to hardware implementations of neural networks. Although the neural network functions correspond to the brain functions, our computer is based on the current technology, which is completely different from the internal structure of the brain based on the neuronal cells. The proposed Reconfigurable Orthogonal Multiprocessor. REOMP, is based on processing units that are reconfigured to execute the algorithms by demanding driven rule. The performance analysis of the architecture is made with the implementation of an artificial neural network, neocognitron, which involves concurrent operations of a great number of artificial neurons. The analysis of the architecture showed that its speed-up is linear in a wide range, where the implementation of REOMP is appropriate. We conclude that the proposed architecture is able to be used to neural network hardware implementation. To obtain the best performance of the architecture, the neural network model should make use of massively parallel neural processing of the previously processed data that is the case of feedforward neural networks.