Matthew Alger1,2, Dr Julie Banfield1, Dr Cheng Soon Ong2,3
1Research School of Astronomy & Astrophysics, The Australian National University, Weston Creek, Australia,
2Data61, CSIRO, Canberra, Australia,
3Research School of Computer Science, The Australian National University, Acton, Australia
Radio host galaxy cross-identification is usually done manually by experts. This will be intractable for large upcoming radio surveys like the Evolutionary Map of the Universe (EMU). Automated cross-identification will be critical for these future surveys, and machine learning may provide the tools to develop such methods. We applied a standard approach from computer vision to cross-identification, sourcing training data from the citizen science project Radio Galaxy Zoo. We used this to cross-identify the 1.4 GHz Faint Images of the Radio Sky at Twenty-Centimeters (FIRST) survey with the AllWISE infrared survey, as well as the 1.4 GHz Australian Telescope Large Area Survey (ATLAS) with the Spitzer Wide-area Infrared Extragalactic (SWIRE) survey. We look at the benefits and challenges of using citizen science like Radio Galaxy Zoo to train machine learning methods for upcoming surveys, and examine the impact of using these complex methods instead of simpler cross-identification rules.
Matthew is a PhD student at the ANU/Data61 interested in applications of machine learning to upcoming large-scale radio surveys to be conducted with SKA pathfinder telescopes.