Optical Associative Processors and Directed Graphs.
Abstract
The thrust of this research concerns associative processors. Many new aspects exist in this effort: large storage capacity, use of storage density as a measure of efficiency, use of new output recollection vector encoding schemes, general 1:1 and pattern recognition many:1 associative, processors, new algorithms and architectures and applications and laboratory realization. We have devised a new higher-order neutral network concept with increased many:one pattern recognition performance. We have also demonstrated a new technique to improve neural network storage capacity and performance by modeling various nonideal components and accuracy during synthesis.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1993
- Accession Number
- ADA268881
Entities
People
- David P. Casasent
Organizations
- Carnegie Mellon University