Digital Simulation of Spiking Neural Networks

Introduction The motivation of digital simulation of artificial neural networks is as diverse as the background of the researchers. Biologists, mathematicians, physicists, psychologists, computer scientists or engineers, all contribute to the field of neural networks. In one case, a very complex model of a single neuron has to be simulated. In another case, a large network of usually simpler neurons must be taken care of. Due to programmability, digital hardware offers a high degree of flexibility and provides a platform for simulations on neuron level as well as on network level. In opposite to signals in nature, digital signals are discrete in value and in time. Therefore, for digital simulations a suitable neuron model needs to be derived. Hence, at first we examine the digital representations of spiking neuron models in subsection 9.2. Then, subsection 9.3 gives an overview of programming environments. The subsequent subsections 9.5-9.7 address the speed-up of pulse-code