A revolution is underway to explore the Cosmic Dawn (CD) and the Epoch of Reionisation (EoR). They are key periods of transition in the early Universe that witness the formation of early structures and the first photon radiating sources. With the constant arrival of ever-richer datasets, observations of the EoR and CD offer a window to an era when the very first stars and galaxies formed and began to alter the large-scale intergalactic medium (IGM) from cold and neutral to hot and ionised. These epochs mark an essential period for understanding astrophysical processes as well as cosmology in the high-redshift Universe.
With ongoing and future experiments, we are set to enter a true ‘big data’ era in high-redshift astrophysics, both with interferometric measurements of the 21-cm signal and observations of radiation from far-infrared, optical, UV and beyond. For instance, radio astronomy is entering a golden era with the help of present and future radio interferometric telescopes, such as the Low-frequency Array (LOFAR), the Murchison Widefield Array (MWA), the Hydrogen Epoch of Reionization Array (HERA) and in particular the Square Kilometre Array (SKA), for the detection of the cosmic 21-cm signal. At the same time, a series of earth- and space-based telescopes such as the Subaru telescope, the James Webb Space Telescope (JWST) and the Euclid telescope aim to unveil the properties of the galaxies present at these times. Moreover, these experiments will be supplemented with observations from the Extremely Large Telescope (ELT) and the Nancy Grace Roman Space Telescope (NGRST) over the next decade.
The wealth of data produced by these observations will propel high-redshift astrophysics and cosmology into the data-driven regime; it is becoming clear that traditional techniques are not up to the challenge of fully exploiting the scale and complexity of these observations. Applications of machine learning methods on big data are able to tackle problems that before had been intractable. Striving to optimally learn physics from our data, our community is enthusiastically adopting these techniques. Therefore it is very timely to survey the landscape of machine learning-based approaches for studying the high-redshift Universe.
For this SAZERAC SIP meeting, we would like to discuss the state and potential of machine learning-based techniques, which can span from statistical to deep learning, in the context of the high-redshift Universe. Key topics include frontiers in the ‘learning’ of EoR and CD observations and simulations in domains such as:
Applications to give a talk are now closed the schedeule has been finalised (below). To participate otherwise you need only sign-up for the general SAZERAC mailing list through which connection info will be shared. Additionally, discussion on the subject is encouraged between the participants through a Slack channel hosted by the SAZERAC team.
You can download the schedule in PDF format here.