Mr Wael Farah1
1Swinburne University Of Technology, Melbourne, Australia
In the last decade, a new class of radio transients, called Fast Radio Bursts (FRBs), have been discovered. FRBs, typically observed as one-off bright millisecond flashes, coupled with hints that they originate from outside our Galaxy, are theorised to be produced by a very energetic mechanism. Thus, FRBs promise to provide insights of potentially new physical processes, sparking the interest of many researchers.
However, discovering FRBs is unarguably a major challenge: the very part of the radio spectrum used for their discovery is contaminated by terrestrial, man-made interference signals (e.g. satellites, mobile phones, electric fences…). Interference can mimic the behaviour of FRBs, sneaking into searched data and producing thousands of false positives that a human has to look though. A solution would be deploying machine learning (ML) based, highly efficient GPU search software that are capable of classifying potential candidates as either interference or genuine FRBs. Moreover, discovering FRBs as-near-to realtime as possible is a key to solving their mystery. Most of the FRBs have been discovered in archival data, years after data recording, enough time for any FRB progenitor to fade away in brightness. The rapid decision making that ML algorithms offer allows the triggering of followup telescopes around the world in a matter of seconds after an FRB has been discovered.
Wael will be presenting an example on how ML has aided in the discovery of FRBs, and what have been learnt after their realtime discovery.
Wael Farah is a PhD candidate in astrophysics at the Swinburne University of Technology. A major part of his thesis is focused on developing novel tools for the discovery of millisecond radio bursts. Wael develops low-latency, machine-learning-based pipelines that segregate genuine astrophysical signals from terrestrial interference. Wael has a masters in astrophysics and B.Sc. in physics.