Optical Neural Nets for Scene Analysis
Abstract
Our objective is to develop new neural net algorithms and architectures for scene analysis and to demonstrate them on a fabricated new hardware laboratory neural net. Our approach marries pattern recognition and neural net techniques and optical/digital technologies. Our hardware laboratory system uses digital and optical neural net hardware in an analog neural net. Our algorithms are intended to be useful on such low accuracy analog hardware. Our algorithms cover a wide range of neural net algorithms and architectures. These can all be utilized on the same laboratory hardware. Our algorithms include five new optimization neural nets (matrix-inversion, mixture multi-target racking, symbolic, and production system neural nets) and an adaptive neural net (adaptive clustering neural net). (jhd)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1990
- Accession Number
- ADA220657
Entities
People
- David P. Casasent
Organizations
- Carnegie Mellon University