Machine Learning for Radio Astronomy

New radio telescopes, such as the SKA, will produce data on such a scale that machine learning approaches for extracting scientific value from those data are essential. However, radio astronomy is a highly specialised field of research and developing machine learning experts with the appropriate domain knowledge is crucial. This project looks at automatic classification and parameterisation of polarized structures in Faraday Depth cubes from radio telescopes. The technique of Rotation Measure Synthesis has revolutionised radio polarization analysis; however, meaningful analysis of the complex structures with the resulting Faraday Depth cubes is still poorly developed. This work will initially use data from the South African SKA precursor telescope, MeerKAT, as part of the MIGHTEE survey with a view to developing techniques that can be deployed for future SKA data analysis.