PATTERN RECOGNITION WITH SELF-ORGANIZING MACHINES
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH SCHOOL OF ENGINEERING
Pagination or Media Count:
Most of the pattern-recognition self-organizing machines can be classified as adjustable-weight threshold-logic machines, statistical-switching machines, or correlation machines. A noisy pattern is a pattern that varies slightly from a model pattern that the machine has been taught. Computer simulations of a typical machine from each of three classes indicate that the noisy pattern recognition capabilities are poor for statistical-switching machines, good for threshold-logic machines, best for correlation type machines. A proposed correlation self-organizing machine is simple, learns in one step, and recog nizes noisy patterns accurately.