STUDY OF LARGE NEUROMIME NETWORKS.
Abstract
The history of neuron modeling is presented and the theory of neuron models is developed within the framework of elementary learning automata. These models consist of three parts: a performance subsystem (an input-output device which has a variable internal state), a performance assessment subsystem, and a conditioning subsystem. The results of three special studies are reported, one in each of the three subsystem areas. The first of these studies deals with the conditioning algorithms used for neuromimes. A random strategy, based on notions from experimental optimization methods, is developed for use with small to medium sized neuromime networks. The second study deals with pulse density codes, their possible mechanizations, and their use in neuron modeling. The third study deals with possible mechanizations of performance assessment subsystems for neuromime networks. A particular objective of this study was the design of a small, expandable network of neuromimes. Many of the problems of large networks are similar to those of small nets and the design effort was carried out to the point of fabrication of a miniaturized Neurotron, the selected neuromime. The Neurotron is a highly versatile neuromime, capable of transforming either analog or digital input signals and providing both algebraic and phase-gain transformations on the input signals. Performance tests of the Neurotron and possible applications of a Neurotron net are presented. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1967
- Accession Number
- AD0824470
Entities
People
- Cary W. Armstrong
- Henry J. Cook
- Lewey O. Gilstrap Jr.