Expand this Topic clickable element to expand a topic
Skip to content
Optica Publishing Group

Parallel Implementations of Neural Networks: Electronics, Optics, Biology

Not Accessible

Your library or personal account may give you access

Abstract

The DARPA report[1] notes that the human cerebral cortex is comprised of 100 billion neurons, each having about 1,000 dendrites that form about 100 trillion synapses. If you multiply that by its operating frequency of about 100 Hz, you get 10,000 trillion interconnections per second. These last two numbers (1014 connections and 1016 connections per second) form the basis for the graph in the DARPA report labeled Computational Requirements. The cortex not only has this prodigious capacity but dissipates about 10 watts, weighs 3 pounds, covers 0.15 square meters and is about 2 millimeters thick. To say it is something substantial for designers of neural systems to shoot for is a considerable understatement.

© 1991 Optical Society of America

PDF Article
More Like This
Electronic vs. Optical Implementations of Neural Networks*

Jay P. Sage
MA2 Optical Computing (IP) 1989

Hybrid optical/electronic implementation of adaptive neural networks

Peter J. de Groot and Robert J. Noll
THHH6 OSA Annual Meeting (FIO) 1988

Learning in Optical Neural Networks

Demetri Psaltis
WA1 Optical Computing (IP) 1991

Select as filters


Select Topics Cancel
© Copyright 2024 | Optica Publishing Group. All rights reserved, including rights for text and data mining and training of artificial technologies or similar technologies.