Development of Free-Space Optoelectronic Neural Systems on Hybrid SI/PLZT Technology.
Abstract
On this grant research was conducted in three directions to make progress towards the goal of implementing large scale learning neural networks on compact optoelectronic multichip modules. (1) Refinement of the Fuzzy ARTMAP neural network algorithm. The algorithm is very well suited to hardware implementation, but previously did not perform well on most real problems. (2) Design, fabrication, and characterization of birefringent computer generated holograms to implement the OTIS optical system for reflective modulators. This optical system provides a k-shuffle interconnection pattern that may be used to implement a backpropagation neural network. (3) Development of a single, compact, efficient optical system (OTIS-VE) compatible with surface-normal transmitters (such as VCSELs) implementing the k-shuffle interconnection pattern.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1996
- Accession Number
- ADA320860
Entities
People
- Sadik Esener
Organizations
- University of California, San Diego