The Cosmostatistics Initiative
The Cosmostatistics Initiative (COIN), an international group built under the auspices of the International Astrostatistics Association (IAA), aims to create an interdisciplinary environment where collaborations between astronomers, statisticians and machine learning experts can flourish.
COIN is designed to promote the development of a new family of tools for data exploration in astrophysics and cosmology.
The group is lead by Rafael S. de Souza (ELTE, Hungary) and co-chaired by Alberto Krone-Martins (Universidade de Lisboa, Portugal), Emille E. O. Ishida (Université Blaise Pascal, France) and Ewan Cameron (Oxford University, UK).
Researchers willing to join are welcome to contact any of the chairs!
The COIN Residence Program
An unstructured and intimate meeting, offering opportunities for collaborative research, learning, and sharing of expertise with concrete scientific outcomes.
The main goal of the CRPs is to benefit young and experienced researchers in astronomy, statistics and computer science alike. Thus contributing to the establishment of Astrostatistics as a discipline on its own.
We also aim at creating a personal bound among our members, provinding a completely informal environment where such intimate connections can grow.
Thus, the program is held in a residence and participants share not only scientific discussions, but participate in logistic decisions during the program.
CRP #1, held in Lisbon 2014 and sponsored by SIM - Universidade de Lisboa, was the first of its kind. We put together a core team working against the clock in order to prepare a series of articles applying modern statistical methods in astronomy.
The researchers worked with approximate Bayesian computation for inference of cosmological parameters, logistic regression for exploration of N-body/hydro cosmological simulations and gamma regression for estimate of photometric redshifts.
Link to the papers, codes an tutorials resulting from CRP #1 are available in the COINtoolbox.
CRP #2 took place in the Isle of Wight - UK, 2015, thanks to the support of Alan Heavens (Imperial Center for Inference and Cosmology - Imperial College of London), Jason McEwen (Mullard Space Science Laboratory - University College of London) and Bruce Bassett (African Institute of Mathematical Sciences, South Africa).
During an intensive week of discussions and coding the participants developed projects related to the application of deep learning for type Ia supernovae spectral characterization, Bayesian logit regression and propensity score matching for the study of AGN/cluster environment, among others.
Results already available are also part of the COINtoolbox.