Description
The SKA will identify tens of millions of new sources - but identifying and classifying their morphologies will require novel, robust, reliable, and efficient methods. In particular, complex sources are of significant interest thanks to (e.g.) wide-angled tails being "weather vanes" for galaxy cluster activity, and the possibility of new discovery space for galaxy evolutionary pathways.
One way of addressing the wealth of data is to use machine learning, and in particular, Self-Organising Maps (SOMs). In this presentation, we detail how we have mapped 251,259 multi-Gaussian complex sources from Rapid ASKAP Continuum Survey (RACS) onto a SOM with discrete neurons. We use Euclidean distance to identify the best-matching neuron for each source and assess the reliability through visual inspection of a subset of sources. We show that sources for which the Euclidean distance to their best matching neuron is ≲ 5 (accounting for approximately 79% of sources) have an estimated > 90% reliability for their SOM-derived morphological labels. Our catalogue of complex radio sources from RACS with their SOM-derived morphological labels from this work will be made publicly available for inspection.
