Accession Number : ADA295046
Title : Dynamical Systems, Neural Networks and Cortical Models.
Descriptive Note : Technical rept. 1 Jul 91-30 Nov 94,
Corporate Author : CALIFORNIA UNIV BERKELEY CENTER FOR PURE AND APPLIED MATHEMATICS
Personal Author(s) : Hirsch, Morris W. ; Baird, William
PDF Url : ADA295046
Report Date : 31 JAN 1994
Pagination or Media Count : 23
Abstract : This work developed new biological models of cortex and produced artificial neural systems for computation which exploited the special capabilities of complex dynamics, and applied them to specific engineering problems: handwritten character recognition, and grammatical inference. Associative memory modules with oscillatory and chaotic attractors were successfully applied to the problem of handwritten character recognition in early stages of the work. In the later stages of the project we succeeded in demonstrating analytically and numerically how an unusual cortical "sensory-motor" computing architecture, can be constructed of recurrently interconnected modules of this type. These learn connection weights between themselves which cause the system to evolve under a clocked "machine cycle" by a sequence of transitions of attractors within the medules, much as a digital computer evolves by transitions of its binary flip-flop states. Because intercommunicating modules of the architecture are analytically guaranteed to store and recall multiple oscillatory and chaotic attractors, the architecture has served as a framework in which to arrange and exploit the spedal capabilities dynamic attractors. Most recently we showed how the architecture can learn to employ selective "attentional" control of synchronization to direct the flow of communication and computation within the architecture to solve a grammatical inference problem. This type of computing architecture and its learning algorithms for computation with oscillatory spatial modes is ideal for implementation in optical systems, where electromagnetic oscillations, very high dimensional modes, and high processing speeds are available.
Descriptors : *NEURAL NETS, *NERVOUS SYSTEM, ALGORITHMS, CONTROL, SPATIAL DISTRIBUTION, OPTICAL EQUIPMENT, HIGH RATE, SIZES(DIMENSIONS), BIOLOGY, MODELS, DYNAMICS, PROCESSING, SEQUENCES, ENGINEERING, SYNCHRONIZATION(ELECTRONICS), COMMUNICATION AND RADIO SYSTEMS, TRANSITIONS, FLOW, ARCHITECTURE, OSCILLATION, LEARNING, ELECTROMAGNETISM, DIGITAL COMPUTERS, CHARACTER RECOGNITION, HANDWRITING.
Subject Categories : Anatomy and Physiology
Medicine and Medical Research
Distribution Statement : APPROVED FOR PUBLIC RELEASE