On Thursday 30 May, the RSS Statistical Computing Section and the RSS Lancashire and Cumbria local group jointly ran a workshop on Machine Learning in Astronomy.
The aim of this event was to bring together statisticians, machine learners and physicists to introduce academics and PhD students to the range of research challenges that sit at the interface of these research fields. There were four speakers and 40 attendees, with a good split across both statisticians and physicists.
Prof David van Dyk (Imperial College London) started the workshop with a talk on 'Data-driven and science-driven Bayesian methods in astronomy and solar physics'. It focused on the interplay of data science, machine learning, Bayesian statistics, data-driven methods, and science-driven methods in the context of several problems in astrophysics, ranging from studying the expansion history of the universe to fitting models for stellar evolution, and mapping the physical characteristics of the solar corona.
Dr Chris Arridge (Lancaster University) gave a talk on 'Research challenges in solar system plasmas'. Chris’s talk introduced a range of problems in the study of planetary magnetospheres, focusing on inference and inversion problems in these systems and the challenges of satellite data collection. Chris’s presentation covered finite wave speed and causal decoupling, as well as time-history effects and the constraints of the governing physical laws.
Dr Florent Leclercq's (Imperial College London) talk on 'Bayesian inference with black-box cosmological models' presented recent methodological advances in likelihood-free inference aimed at fitting cosmological data with 'black-box' numerical models. His talk focused on Bayesian optimisation and the Taylor-expansion of the simulator as two different solutions to address the challenges of analysing large-scale astronomical surveys. Florent discussed how advancing this research frontier leads to unique statistical problems when trying to unlock the information content of massive and complex data streams which can inform on the origin and evolution of the Universe.
The final speaker was Dr Ingo Waldmann (UCL) who spoke about 'Deep learning exoplanets and the solar system'. Ingo gave an overview of the current state of machine learning in the field of extrasolar planetary physics and provided a showcase of two recent examples of solving long-standing problems with machine learning: 1) The modelling of exoplanetary atmospheres using a new deep learning framework, ExoGAN and 2) The mapping of Saturn’s storm regions using a new hyperspectral image classification approach, PlanetNET. The focus of the talk was on the development of intelligent algorithms and how they play an important part in facilitating the analysis of large and rich data sets.
The event concluded with a poster session, where PhD students from physics and statistics presented their work over a wine reception. This was a great opportunity for local statisticians to engage with physicists, and we hope that some future collaborations will emerge from this highly informative workshop.