MODEL OF AN ASYNCHRONOUS RESONANCE PROCESSOR

Period of Performance: 01/01/1993 - 12/31/1993

$37.3K

Phase 1 SBIR

Recipient Firm

Parallel Logic Inc.
252 Brickyard Rd
Farmington, CT 06032
Principal Investigator

Abstract

DUE TO THE NEED FOR FASTER PROCESSING, A GROWING NUMBER OF NEWER COMPUTING MODELS ARE INCREASING SPEED WITH INTRIGUING PHYSICAL DESIGNS. MASSIVE PARALLELISM, FOR EXAMPLE, EXPLOITS UNIQUE PHYSICAL ARRANGEMENTS IN ORDER TO INCREASE COMPUTER PERFORMANCE. SEEMINGLY, THE SIGNAL ITSELF IS BECOMING LESS IMPORTANT. ONE CAN EXTRAPOLATE FURTHER AND ENVIRISION A COMPUTER CONCEPT, WHERE THE ONLY ROLE OF THE SIGNALS IS TO ESTABLISH A STABLE STRUCTURE, WHICH IN TURN WOULD IMPLY SOLUTION. THE OBJECTIVE OF THE RESEARCH IS TO EXAMINE A WORKING HYPOTHESIS THAT CERTAIN HIGH PERFORMANCE COMPUTING CAN BE DONE BY NEGOTIATING EQUILIBRIA OF APPROPRIATE ANTI-SYMMETRIC DIGITAL CIRCUITS THROUGH RANDOM NOISE. THE EXPERIMENTAL CIRCUITS ARE EMBODIMENTS OF PARTICULAR ALGORITHMS. IF THE PARTIAL INPUTS AND OUTPUTS KEEP WELL-DEFINED VALUES, RESEARCHERS BELIEVE THAT IT IS POSSIBLE TO SUPERIMPOSE CHAOTIC PERTURBATIONS IN SUCH A WAY THAT THE PROCESSOR WILL SETTLE TO ITS STABLE STATE IF, AND ONLY IF, IT ATTAINS A SOLUTION OF THE ALGORITHM. SUCH AN ASYNCHRONOUS RESONANCE PROCESSOR CAN BE A HIGH PERFORMANCE PLATFORM FOR A VAREITY OF APPLICATIONS PRIMARILY IN THE AREA OF MACHINE INTELLIGENCE.