Monolithic Optoelectronic Implementation of Neural Planes
Abstract
The implementation of a neural computing system, whose structure and function is motivated by natural intelligence, will provide a unique way to solve problems that are typically too difficult for the conventional electronic computers to tackle. Interest in this type of computer has emerged largely because it is hoped that by building a computer that share some of the characteristics of the biological systems, we will be able to address problems such as image recognition which animals do exceedingly well but current computer do not. There had been considerable progress on the theoretical research in neural network to justify the optimism of future applications. This has resulted in a focused attention on the hardware realization of neural architectures. The computational power of neural computer arises from matching the computer architecture and the physical properties of the devices used in the implementation to the requirements of the problem. In other words, a neural computer is highly specialized and it is therefore very difficult to derive its full advantages on a general purpose computer. This provides a strong impetus for advancing the technologies for the physical realization of neural computers in parallel and interactively with the development of theoretical neural network models.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1991
- Accession Number
- ADA268915
Entities
People
- Annette Grot
- Demetri Psaltis
- Steven Lin
Organizations
- California Institute of Technology