An experimental analog VLSI neural network with on-chip back-propagation learning
Analog Integrated Circuits and Signal Processing
| Hi-index | 0.00 |
Today Feed Forward Neural Networks (FFNs) use paradigms tied to mathematical frameworks more than to actual electronic devices. This fact makes analog neural integrated circuits heavy to design. Here we propose an alternative model that can use the native computational properties of the basic electronic circuits. A practical framework is described to train analog FFNs in compliance with the model. This is especially useful whenever the weight storage elements cannot be re-programmed on the fly at a high rate. To show how the capability of such framework can be applied to neural systems with non conventional architectures two cases are presented. The first one is a neural signal processor named NESP which has sigmoidal neurons and the other is an innovative architecture named N-LESS.