Neuromime Networks for Multiprocessors.
Abstract
Multiprocessors are systems that handle many inputs and outputs simultaneously. This report is concerned with the use of neuromime nets in multiprocessors. Large neuromime nets contain from 1 million to 1 billion neuromimes. The conceptual framework of large nets is reviewed and the major problem areas of large nets are described. The high volume of information that can be handled by large nets renders input and output an intractable problem unless steps are taken to organize these information flows. A technique, termed parameter estimation, for handling image input information is developed and an approach to the formal description of outputs of complex automata is described. These techniques are employed in a discussion of the functions required of autonomous and semi-autonomous sytems. Goal structures and learning in such systems are discussed. Random searches, which are very useful in adaptive systems theory, are reviewed in detail. Possible hardware realization of large neuromime nets, through the use of multiplex techniques, is discussed and design concepts presented. System design and utilization are considered for possible applications. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1968
- Accession Number
- AD0842467
Entities
People
- Albert V. M. Ferris-prabhu
- Lewey O. Gilstrap Jr.