School of Chemistry & Physics

Dr Chandan Nagarajappa.

Machine Learning Used to Deepen Understanding of the Cosmos

Dr Chandan Nagarajappa has graduated with a PhD in Physics for his dissertation titled: Application of Machine Learning in Cosmology Computer Literacy in Banach Spaces.

Supervised by Professor Yin-Zhe Ma, Nagarajappa’s research focused on the interface of machine learning (ML) and cosmology, in particular, using machine learning to study the primordial non-Gaussianity (PNG) of the cosmic microwave background radiation.

‘I am very glad that Chandan has been awarded his PhD as the journey to completion was very rocky,’ said Ma. ‘He came to UKZN in late 2019 and, as we all know, from 2020 the world experienced more than two years of very unusual lockdown and restricted communications. Throughout this whole process, even though he was a foreign student far away from his home, Chandan persisted with his research, eventually obtaining very good results that were published in the international peer-reviewed journal, the Monthly Notices of the Royal Astronomical Society.

‘The PNG, produced during the cosmic inflation period (the fast expansion of the Universe), can arise as the coupling of long-wavelength fluctuations and short-wavelength perturbations,’ said Ma. ‘Chandan developed a machine learning tool to recognise such “correlation patterns” in the cosmic microwave background map. By employing cutting-edge machine learning techniques, he provided an alternative to probe the initial condition of the Universe and the cosmic evolution. This can be regarded as a very novel approach to the non-Gaussianity study in cosmology.’

Said Nagarajappa: ‘At this intersection of machine learning and cosmology, machines can be trained in different aspects of Astro needs like pattern recognition, classification, RFI detection, predicting properties and data compression.

‘I use many ML techniques which serve as alternate ways to solve problems in astrophysics and cosmology, using these computer algorithms to analyse and understand the large amounts of data related to the study of the universe’s structure, origin and evolution.

‘In simple terms, it’s like teaching a computer to learn from the information it receives about the cosmos, just like how we learn from our experiences.’

He said the significance of his research was that using ML techniques enabled researchers to bypass traditional techniques that require heavy resources.

Nagarajappa has always been intrigued by the vastness of our cosmos: ‘It is both fascinating and humbling, filled with countless unanswered questions and unexplored mysteries. The formation and evolution of the universe are complex processes that continue to captivate scientists and researchers worldwide and it has a huge impact on my thinking. Just to imagine the scale of the universe in my mind becomes impossible. That’s’ where the interest to study cosmology comes in. It is my way of being able to look at the sky, understanding its scale and its beauty while accepting its mysteries.’

Nagarajappa, from Karnataka in India where he gained his MSc at the University of Mysore, chose to pursue a PhD at UKZN because of its strong astro group and exciting astrophysics projects as well as its positive community environment. ‘These factors contribute to a stimulating and supportive atmosphere for one’s academic and personal growth,’ he said. ‘It is always helpful to discuss your work in a community. Even when the fields are different a lot of understanding and ideas are generated while explaining your work.’

He thanked his supervisor Professor Yin-Zhe Ma for his guidance and assistance with funding; as well as the South African Theory School (SATS), which initiated online courses during the COVID-19 pandemic – he found Professor Jonathan Shock’s course on ML particularly helpful when conducting his research.

Nagarajappa hopes to take up a postdoctoral fellowship position and build a career in academia. ‘I am passionate about exploring the enigmatic field of cosmology, while also considering its intersection with innovative technologies like machine learning,’ he said.

His advice to fellow researchers: ‘Make sure to be in a good research group which is active.’

Words: Sally Frost

Photograph: Supplied