A novel massively associative processing architecture for the implementation of artificial neural networks

The author presents the ASP (associative string processor) a massively parallel, programmable, fault-tolerant architecture, which can efficiently support low-MIMD/high-SIMD and other parallel computation paradigms by providing comprehensive cover of both numeric and symbolic processing, supporting content-addressing, and a dynamically reconfigurable rich-in-connectivity interprocessor communication network. The efficient implementation of a neural network model (i.e. the backpropagation model) on the ASP is addressed and its performance, at a computational rate of 10/sup 9/ interconnections per second (i.e. 10/sup 9/ Giga-interconnections/s), is reported.<
>