Modelling neural dynamics on neuromorphic hardware

Neuromorphic engineering has allowed the production of biologically-inspired hardware on which large-scale Spiking Neural Networks (SNNs) can run. SNNs are a newer generation of the extremely common and successful Artificial Neural Networks (ANNs) which are typically used in machine learning and artificial intelligence tasks. While ANNs have been successful, they rely heavily on vast amounts of computing power and cloud computing; despite the name they are not accurate models of biological systems. A human brain - which is a SNN - is able to complete more different and complex tasks than any moderately sophisticated ANN whilst using orders of magnitude less energy in the process.
Artificial SNNs are a new generation of ANNs which are biologically-inspired and consist of accurate models of individual neurons in large numbers. Neuromorphic hardware such as SpiNNaker has allowed simulation of large SNNs but, thus far, these networks are based on quite abstract neural models. As models are developed, computational neuroscientists are becoming increasingly convinced that more sophisticated modelling of the biology may be needed if brain-like functions are to be duplicated.
This project aims to develop more complex models of neural dynamics using the SpiNNaker neuromorphic computer, where software allows experimentation with the various details which may be included. One early example will be the incorporation of models of individual ion channels within neural membranes and exploring their impact on large-scale SNNs.