Page 46 - University of Pretoria RESEARCH REVIEW 2018

Harnessing vast data and scientific discovery
Christopher Cleghorn, Department of Computer Science
The MeerKAT telescope is a great opportunity for South Africa to demonstrate its scientific prowess. However, given the immense data generated by the telescope, the most rapid and scalable way to capitalise on this opportunity is via machine learning.
crucial for algorithmic effectiveness in the real world. In 2018, he published an overarching stability theorem
for a broad class of particle swarm optimisation algorithm variants. The result obtained relied on provably minimal modelling assumptions and the theorem marks a unification of previous work on stability in the field.
The more recent radio astronomy research focus has been borne out of a CIRG collaboration with UP’s astronomy group in the Department of Physics. The ASTRO-CIRG group aims to harness the full power of AI to accelerate the advancement of fundamental astrophysics. To make this collaboration truly effective, ASTRO-CIRG has physicists and computer scientist working in an integrated lab environment.
The initial focus of the ASTRO-CIRG group is two-fold: first, to build and train optimal goal-orientated image stacking controllers, a method already used to detect previously unknown physical phenomena. The second focus is on constructing machine learning approaches to perform rapid morphological classification, in the hope of not only cataloguing, but
also making possible the detection of previously unseen morphological structures.
The use of Artificial Intelligence (AI) techniques will be of paramount importance to ensure scientists are able quickly to detect potentially subtle, astronomical phenomena.
Led by Dr Christopher Cleghorn
in the Department of Computer Science has historically focused on fundamental research in the area of stochastic optimisation and neural network-based machine learning. In light of the massive potential of the MeerKAT telescope, Dr Cleghorn has pivoted CIRG’s research to ensure the
group is able to harness expertise in machine learning to facilitate scientific discovery in astrophysics.
There are currently two primary research focus areas in the field of AI: the theoretical analysis of population- based optimisation algorithms, and the application of machine learning techniques to radio astronomy.
In terms of the former, Cleghorn’s focus has been on the mathematical derivation of stability criteria for optimisation algorithms. Stability is
a facet of algorithm dynamics that is
Christopher Cleghorn and Roger Deane.
44