 |
Custom Designed Electro-Optic Components for Optically Implemented, Multi-layer Neural Networks,
Authors: M. G. Robinson; K. M. Johnson; D. Jared; D. Doroski; S. Wichart; COLORADO UNIV AT BOULDER DEPT OF ELECTRICAL AND COMPUTER ENGINEERING |
|
Abstract:
Optical implementations of one-layer, perceptron-like neural networks have been shown to be very successful at associating pattern/target sets despite large system errors. It has shown that large systems can be realized with such architectures (> or = 40,000 interconnections), and appreciable processing speeds have been demonstrated (> 10 to the 8th power interconnections/sec). However, single layer networks are limited due to their inability to associate patterns that are not linearly separable. A more general network is the two-layer network, which is able to model arbitrary functions, and creat any decision boundary within the input vector pattern space. In order to implement such a network, it is necessary to perform a nonlinearity at the hidden layer before performing a subsequent matrix multiplication. In general, optical materials performing fast nonlinear processing require high optical powers. Hybrid opto-electronic devices can perform nonlinear operations at moderate speeds and low optical powers.
| Pages: |
4 |
| Report Date: |
22 MAY 1992 |
| Report Number: |
P026800 |
Report Unavailable |
| This title is unavailable from Storming Media. We do not know when it might be available, if at all. We list the report on our site for bibliographic completeness, to help our users know what other work has been performed in this field. Please note that as with all titles on this site, we do not have contact information for any of the authors. Nor can we give any suggestions on how one might obtain this report. |
|
|
|
|
|
|
| |
|
|
Keywords relating to this report:
Email This Abstract
