An Analog VLSI Implementation of Hopfield's Neural Network (1989)
| Citations: | 7 - 0 self |
BibTeX
@MISC{Verleysen89ananalog,
author = {Michel Verleysen and Paul G. A. Jespers},
title = {An Analog VLSI Implementation of Hopfield's Neural Network},
year = {1989}
}
Bookmark
 |
 |
 |
 |
 |
 |
OpenURL
Abstract
ective simulations of neural networks exceed the limits of conventional machines, researchers have been working on implementations that are more adapted to inherent neuronal properties. After a short discussion of fixed-value and digital networks, we point out the advantages of analog circuits for neural networks. Electronic implementations We can characterize electronic neural networks by their degree of connec- tivity. Hopfield's networks 2 interconnect fully. A feedback process connects the output of each neuron to the input of each other neuron. In Rumclhart's multilayer Percepttons, the output of each neuron connects to all neurons in the next layer--but no connection occurs within a layer. In locally interconnected networks, neurons connect only to their nearest neighbors. An example of this type of network is the silicon retina implemented by Mead? 46 IEEE MICRO 0272-1732/89/1200-0046501.(}0 ( 1989 IEEE In2 i rl n ////' w n Threshold function Out, Figure 1. Structure of a
