Presenter: Kevin Heng, University of Bern, Center for Space and Habitability
Title: Interpreting Spectra from Exoplanets Using Bayesian and Machine Learning Methods
Exoplanet science is in the midst of an inexorable revolution, where detection of new objects has become routine and the entire community is shifting towards the characterization of their atmospheres. Interpreting spectra of these exo-atmospheres is potentially a powerful way of asking questions about formation history and habitability conditions. In the current talk, I focus on how these spectra are interpreted using both traditional Bayesian methods (nested sampling) and supervised machine learning (random forest). The random forest method may be understood as being part of Approximate Bayesian Computation (ABC) methods, although I continue to have doubts about how formal or rigorous the relationship is. Nevertheless, I demonstrate that the random forest offers powerful analysis/interpretation advantages not easily accessible to traditional Bayesian methods. If time permits, I will discuss a cutting-edge problem of combining low- and high-resolution spectra using essentially a form of physically-motivated “summary statistics”, which when combined with the random forest method produces a novel approach for interpreting spectra with 10 million data points or more. There is plenty of fertile ground for interdisciplinary collaboration between astrophysicists, data scientists and machine-learning experts in the era of the James Webb Space Telescope and next-generation, ground-based high-resolution spectrographs.
As usual, the format will be a short presentation followed by plenty of discussion.
Please let all interested know about the Journal Club (get them to send their Email address). We are looking forward to lively discussions.