Optically Connected Multiprocessors for Simulating Arti cial Neural Networks

This paper investigates the architectural requirements in simulating large neural networks using a highly with distributed memory and optical interconnects. First, we model the structure of a neural network and of individual cells. These models are used to estimate the volume of messages that need to be exch processors to simulate the weighted connections of the neural network. The distributed processor/memory to an electronic implementation for greater versatility and exibility. Optical interconnects are used to sa communication bandwidth demands. The hybrid implementation attempts to balance the processing, m demands in simulating asynchronous, value-passing models for cooperative parallel computation with self