Higher Degree by Research Application Portal

This is an opportunity for a PhD in a rapidly growing area in Machine Learning. Spiking neural networks (SNNs) much more closely mimic the operation of a real brain than traditional artificial neural networks (ANNs) by reproducing the spike-driven nature of real neurons. There is also the opportunity to collaborate with SynSense AG, a startup developing neuromorphic hardware to execute SNNs at low power.

While most simulations of spiking neurons networks (SNN) are conducted with 32-bit floating point precision, digital spiking neuron hardware often operates with quantised integer logic, at reduced bit depths.

SNNs have several computational requirements for spiking neurons, which should be aligned with efficient hardware implementation. The first is simulation or approximation of exponential decay, to support synaptic and membrane dynamics. Accuracy here is required to ensure compatible dynamics between floating-point simulations and hardware implementation. A HW numerical representation that makes exponential decay both cheap and accurate would be highly desirable.

The second is a requirement for accurate representation of numbers close to threshold. Whether or not a neuron crosses threshold has a large impact on the performance and behaviour of a spiking network. We suggest an exploration of whether concentrating the numerical representation around threshold (e.g. around 1.0) provides benefits to network accuracy or simulation accuracy in the face of reduced bit-depth. 

Please contact Andreas Wicenec in the first instance.

The ideal candidate will hold either a Mathematics, Computer Science or Electronic Engineering Bachelor Honours or Masters degree with some experience in computational methods and design for embedded hardware.

Skills

• Python

• Machine Learning optimisation. Experience with DNN training is desirable 

• Low-level HDL coding

• Knowledge of numerical representations is desirable

Training and Development

The candidate will be responsible for designing and evaluating numerical representations, in the context of SW and HW simulations of spiking neurons. Potential collaborations with SynSense, a commercial neuromorphic processor hardware startup, are available. 

The Doctor of Philosophy (PhD) is a program of independent, supervised research that is assessed solely on the basis of a thesis, sometimes including a creative work component, that is examined externally. The work presented for a PhD must be a substantial and original contribution to scholarship, demonstrating mastery of the subject of interest as well as an advance in that field of knowledge. 

Visit the course webpage for full details of this course including admission requirements, course rules and the relevant CRICOS code/s.