VHDL Implementation of Neurone Networks Using a Simplified Action Potential Waveform

This paper describes a model of a neurone and a network of neurones based on their biological characteristics observed in their basic input and output relationships. Previous work has postulated that artificial neurones that are more biological in their function can form more robust and noise-tolerant neurone networks compared with conventional artificial neural networks (ANN). Existing neurone models are computationally intense because they are based on complex biological processes relating to biochemical and electrophysiological descriptions of neurone operations. Consequently building a neurone network by connecting these types of neurones becomes impractical. Therefore, a digital model of a neurone has been developed using the hardware description language VHDL. This digital neurone model was designed to be simple so as to be fast and not require large hardware resources. Although it is simple, it is biologically plausible because it uses a digitised action potential waveform and randomised 3D growth based on the properties of biological neurones. The network of digital neurones can then be synthesised into digital hardware using a commercial synthesis tool.