Accession Number : AD0655650
Title : ELECTROCHEMICALLY ACTIVE, FIELD-TRAINABLE PATTERN RECOGNITION SYSTEMS.
Descriptive Note : Final rept. 1 Feb 66-30 Apr 67,
Corporate Author : SPACE-GENERAL CORP LOS ANGELES CALIF CENTER FOR RESEARCH AND EDUCATION
Personal Author(s) : Stewart,R. M. ; Milne,J. R. ; Hickey,G. I. ; Hendrix,C. E.
Report Date : APR 1967
Pagination or Media Count : 84
Abstract : Research was directed toward the development of a new class of 'pattern recognition' equipment for rapid and automatic detection and/or classification of spatio-temporal patterns. One restricted but key class of systems was emphasized--functionally, adaptable linear decision functions; structurally, a homogeneous array of gold/iron dipoles in nitric acid which feed, in parallel, a low-impedance detector circuit--and means for direct excitation of such arrays by patterns of incident light. Theory shows that such a system (and some others) can be 'trained' through the agency of 'diffuse' or 'global' electrical shocks applied across the whole array immediately following each improper response in a training sequence and, thus, avoid the conventional requirements for detailed access, connecting, or 'internal' structuring and programming. During this process the size and structure (and, hence, function) of the dendrites is gradually altered in excited regions through differential dendritic electrodeposition and dissolution. When perfected and extended, this type of 'bulk' process and others like it should lead to pattern recognition machines reaching entirely new levels of packing density and versatility. Additional design data was acquired through a number of basic experiments and completed an experimental sixteen-element 'Linear Field-Trainable' (LIFT) array was completed together with compatible pattern input, control, and monitoring equipment.
Descriptors : (*PATTERN RECOGNITION, ELECTROCHEMISTRY), (*ELECTROLYTIC CELLS, PATTERN RECOGNITION), (*ADAPTIVE SYSTEMS, ELECTROLYTIC CELLS), IRON, GOLD, FILAMENTS, DENDRITIC STRUCTURE, DETECTORS, ELECTRODEPOSITION, NITRIC ACID, LEARNING MACHINES
Subject Categories : Physical Chemistry
Distribution Statement : APPROVED FOR PUBLIC RELEASE