ANALOG NEUROMORPHIC COMPUTATION - AN APPLICATION TO COMPRESSION

Nature has evolved computing engines whose intelligence and natural ab ilities are unrivaled by modem computers. To match Mother Nature's abi lities, we must overcome the same difficulties faced by natural system s, and we must learn to perform reliable computing with unreliable com ponents. Steps in this direction are being taken by several groups at the Applied Physics Laboratory and at The Johns Hopkins University Dep artment of Electrical and Computer Engineering. The purpose of the wor k is twofold: (1) to explore algorithms based on physical and neural m odels of computation and (2) to develop useful applications. We descri be the basic approach and an experimental electronic neural network fo r the decompression of one-dimensional signals.