Letters of Intent received in 2018
Non-GA Symposium: Bayesian Inference in Astronomy
||27 July 2020 to 31 July 2020
||Imperial College London, United Kingdom
||Daniel Mortlock (firstname.lastname@example.org)
||Division B Facilities, Technologies and Data Science
Chair of SOC:
||Daniel Mortlock (Imperial College London)
Chair of LOC:
||Andrew Jaffe (Imperial College London)
Cosmology (hierarchical parameter inference; model comparison; consistency tests)
Astronomical surveys (image analysis; data reporting; confusion/blending)
Gravitational waves (parameter inference; cross-identification; sky localisation)
Galaxies (morphology, photometric redshifts, non-parametric models)
Exoplanets (population inference; time series analysis; model comparison)
Stars (hierarchical parameter inference, combining multiple constraints)
Methods (model testing/checking, sampling techniques, big data)
Modern astronomy is an almost completely quantitative and data-driven branch of science; it is also purely observational, with minimal freedom to design experiments. As such, it is a perfect match for the techniques of Bayesian inference, a fact which has been implicitly acknowledged by the exponential growth in the use of Bayesian methods over the last few decades. Bayesian inference is the default approach in some fields (e.g., cosmology) and is becoming increasingly popular in many others (e.g., exoplanets). The increasing importance of statistics in astronomy is illustrated by the inclusion of an astro-statistics working group in the International Statistical Institute, (roughly the statistical counterpart of the IAU), and later, in 2012 an IAU Working Group in Astrostatistics and Astroinformatics was created. The IAU then had its first Symposium dedicated to astro-statistics (Statistical Challenges in 21st Century Cosmology in Lisbon in 2014). The field has now grown to the point that there is a sufficiently large sub-community of Bayesian astronomers to warrant an IAU Symposium dedicate to this topic, and we propose to host the first such meeting, titled Bayesian Inference in Astronomy, in London in July 2020.
The most obvious group who will be served by this meeting are those who are already successfully using Bayesian inference to obtain exciting and rigorous results. The focus will be on science results obtained using Bayesian methods - particularly in cases where Bayesian inference was the only route to obtaining a reliable result - and the list of topics above is hence focused on science areas, not methodology. The science areas identified above reflect those in which Bayesian inference is most common, but are not exhaustive.
The second main aim of this meeting is to serve both statisticians and astronomers who are making methodological and algorithmic developments that enable Bayesian methods to be utilised for an increasing set of problems which could not previously be handled in such a principled manner. This includes the development of Bayesian model-testing and model-checking formalisms (particularly relevant to purported anomalies in the cosmological model), sampling and simulation algorithms (e.g., nested sampling, Hamiltonian Monte Carlo, sparsity-based methods). The lesson from the massive take-up of Bayesian methods in astronomy is that what was impossible only a few years ago is, thanks to algorithmic and computing developments, readily achievable. As such we also hope that the meeting will be of use and interest to astronomers who would like to adopt Bayesian methods themselves, and we will host some break-out sessions to facilitate this.