Optical Computing Based on Neuronal Models.
Abstract
Ever since the fit between what neural net models can offer (collective, iterative, nonlinear, robust, and fault-tolerant approach to information processing) and the inherent capabilities of optics (parallelism and massive interconnectivity) was first pointed out and the first optical associative memory demonstrated in 1985, work and interest in neuromorphic optical signal processing has been growing steadily. For example, work in optical associative memories is currently being conducted at several academic institutions (e.g., California Institute of Technology, University of Colorado, University of California-San Diego, Stanford University, University of Rochester, and the author's own institution the University of Pennsylvania) and at several industrial and governmental laboratories (e.g., Hughes Research Laboratories - Malibu, the Naval Research Laboratory, and the Jet Propulsion Laboratory). In these efforts, in addition to the vector matrix multiplication with thresholding and feedback scheme utilized in early implementations, an arsenal of sophisticated optical tools such as holographic storage, phase conjugate optics, and wavefront modulation and mixing are being drawn upon to realize associative memory functions.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1987
- Accession Number
- ADA191481
Entities
People
- Nabil H. Farhat
Organizations
- Moore School of Electrical Engineering