Accession Number : ADA293111
Title : Neurodynamical Systems for Cognition and Target Identification.
Descriptive Note : Final technical rept. 1 Jul 91-30 Jun 94,
Corporate Author : PENNSYLVANIA UNIV PHILADELPHIA
Personal Author(s) : Farhat, N. H.
PDF Url : ADA293111
Report Date : OCT 1994
Pagination or Media Count : 83
Abstract : Our study of cognitive automated target recognition based on the neural paradigm for information processing reveals that inclusion of bifurcation and synchronicity (phase-locking) in network dynamics can markedly improve the performance of ATR systems. This gave impetus to our study of how synchronicity could arise in cortical networks when it is known the brain has no central clock. Raising this question has led us, through analysis of models of biological neurons employing the tools of nonlinear dynamics, to the development of the bifurcating neuron concept and model. This spiking neuron model combines functional complexity comparable to that of biological neurons with structural simplicity and low power consumption when implemented electronically or optoelectronically. These attributes make the bifurcating neuron ideally suited for use as building block of a new generation of spiking neural networks that employ phase-locking, bifurcation and chaos, on the single processing element level, to emulate higher-level cortical functions such as feature-binding and cognition that are essential for advanced ATR systems, and other operations like separation of object from background, inferencing and rudimentary reasoning.
Descriptors : *NEURAL NETS, *COGNITION, *TARGET RECOGNITION, *SYSTEMS ANALYSIS, *INFORMATION PROCESSING, LOW POWER, AUTOMATION, BIOLOGY, BRAIN, CLOCKS, MODELS, TOOLS, DYNAMICS, REASONING, PROCESSING EQUIPMENT, IDENTIFICATION, NONLINEAR SYSTEMS, LIMITATIONS, NERVE CELLS, SEPARATION, SYNCHRONISM, ENERGY CONSUMPTION, PHASE LOCKED SYSTEMS, BIFURCATION(BIOLOGY).
Subject Categories : Computer Systems
Distribution Statement : APPROVED FOR PUBLIC RELEASE