[Skip to Content]
Example of a Self-Organised Map used to classify radio sources by their morphology (Polsterer et al. 2014).

Example of a Self-Organised Map used to classify radio sources by their morphology (Polsterer et al. 2014).

The next generation of radio telescopes will greatly expand the number of radio sources we know by several orders of magnitude. Even with increased computing power and algorithms it will not be possible to classify them all, hence new techniques must be found. The burgeoning field of Machine Learning provides powerful and novel tools to use in identifying radio sources including kth Nearest Neighbour and Self-Organised Map amongst others.

These methods will be expanded and developed to tackle several problems in the radio surveys including:

  • Identification: cross-matching a radio source with surveys at other wavelengths, which is a key requirement to understanding the radio source.
  • Classification: determining whether the radio source is powered by star formation or black hole activity including automatically classifying the radio morphology.
  • Distance estimation: usually distances are determined from spectroscopy, which is expensive in terms of telescope time, but this project will also refine methods to obtain statistical distance estimates from more limited data.

This project will be at the fusion of astronomical observations and modern computing problems requiring a student with interest in both aspects.

This project will develop tools to be tested on and used for the Evolutionary Map of the Universe (EMU http://emu-survey.org/) on the Australian Telescope Square Kilometre Array Pathfinder. This survey is expected to find around 70million radio sources including most of the star forming galaxies in the Universe and the first black holes. EMU is an international project with over 200 members from around the world.

Co-Supervisor

Dr Paul Hancock

Early Career Research Fellow, Space Detective

Read More