Compact building blocks for artificial neural networks
Overview
Identity
Additional Document Info
Other
View All
Overview
abstract
A four quadrant analogue multiplier, an activation function and a differentiator intended for the implementation of backpropagation algorithms are proposed. The priorities for the design are low-power, low-voltage, and minimum silicon area. Breadboard results, showing good agreement with the theoretical ones, are reported. IEE 1999.