Artificial neural network on a SIMD architecture

An implementation of a fully connected artificial neural network using the multilayered perceptron model is described. The neural network is implemented on a systolic array processor based on the Geometric Arithmetic Parallel Processor (GAPP) chip. Arrays of GAPP chips make up a single-instruction multiple-data (SIMD) class machine which has fine-grained connections and is fully programmable. Previous application areas of the GAPP system are image/signal processing, computer vision, and knowledge-based processing. The neural network is a relatively new processing model for the GAPP, but one that readily maps onto the architecture of the overall array processor. The proof-of-concept neural network is a multilayered perceptron model which uses the back-propagation learning paradigm. This initial network has fewer than 100 nodes in three layers and is trained to recognize letters of the alphabet