Optical Computing Based on Neuronal Models
Abstract
The ultimate goal of the research work carried out under this grant is understanding the computational algorithms used by the nervous system and development of systems that emulate, match, or surpass in their performance the computational power of biological brain. Tasks such seeing, hearing, touch, walking, and cognition are far too complex for existing sequential digital computers. Therefore new architectures, hardware, and algorithms modeled after neural circuits must be considered in order to deal with real-world problems. Neural net models and their analogs represent a new approach to collective signal processing that is robust, fault tolerant and can be extremely fast. These properties stem directly from the massive interconnectivity of neurons (the logic elements) in the brain and their ability to perform many-to-one mappings with varied degree of nonlinearity and to store information as weights of the links between them, i.e., their synaptic interconnections, in a distributed non-localized manner. As a result signal processing tasks such as nearest neighbor searches in associative memory can be performed in time durations equal to a few time constants of the decision making elements, the neurons, of the net.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1988
- Accession Number
- ADA197749
Entities
People
- Nabil H. Farhat
Organizations
- Moore School of Electrical Engineering