Letters of Intent for 2014

LoI 2014-164
Statistical Challenges in 21st Century Cosmology (SCCC 21)

Date: 19 June 2014 to 23 June 2014
Location: Lisbon, Portugal
Contact: Alan Heavens (a.heavens@imperial.ac.uk)
Coordinating division: Division B Facilities, Technologies and Data Science
Co-Chairs of SOC: Alan Heavens (Imperial College, London)
Jean-Luc Starck (Commissariat à l’Energie)
Chair of LOC: Alberto Krone-Martins (Lisbon University)



• Cosmic microwave background
• Weak lensing
• Large-scale structure
• High redshift supernovae.
• Mapping high-z 21-cm radiation
• Lyman-alpha forest
• Astronomical discovery from overwhelmingly large datasets.
• Statistical methods used in astronomical data analysis (including new developments coming from fertile cross-interactions in astrostatistics). A preliminary list of these methods is included below:
- Bayesian methods, model selection.
- Multivariate classification,clustering.
- Sparsity: wavelets, compressive sampling, 2D/3D data representations.
- Machine learning for large multivariate datasets: Kernel regression, Support Vector Machine, neural networks, supervised learning.



During the past 20 years recent years there has been a resumption of a dialogue between astronomers and statisticians, led by the Penn State conferences organized by Professors Jogesh Babu and Eric Feigelson. This dialogue has been fruitful and has been the origin of a new discipline that can be called Astrostatistics. The main tools for comparing theoretical results with observations in astronomy are statistical. However, the development of huge astronomical databases presents challenges of scale, and has initiated an active use of newly-developed statistical techniques in astronomy, notable examples being sparsity and compressed sensing. The meeting is especially timely from the point of view of cosmological surveys, where the size makes application of a fully Bayesian analysis computationally extremely demanding, especially in the realm of model selection. As examples, in the field of the microwave background, Planck’s resolution leads to a dataset whose size is so large that analysis even at the two-point level is non-trivial, and at higher-order extremely challenging; Pan-STARRS1 will have a complete survey of 3pi steradians of petabyte size; the Dark Energy Survey and the VST KiDS surveys will be well underway and offering similar difficulties of analysis, and the cosmological community will be preparing for LSST and for Euclid, a survey of a large fraction of the sky at a resolution close to that of the Hubble Space Telescope. Wide-field spectroscopic cosmology surveys of will be contemplating surveys of over 10 million objects with a spectral resolution of 5000, and the SKA precursors will be grappling with data challenges which currently are unsolved. These examples also highlight the big current role and even bigger future role of archival data in astrophysics research, and the meeting offers an opportunity to show-case techniques and methodologies that will need to be used by the wider community to use these archival data. There have not yet been any IAU symposia devoted to Astrostatistics; such a symposium, with a special scientific focus on cosmological inference, is timely and of great interest.

Astrostatistics (sub)organizations within the ISI and IAU have been very active in the recent years. The International Statistical Institute, (roughly) the counterpart of the IAU, started an Astrostatistics Committee in 2009 under the leadership of statistician Joseph Hilbe. Now being reorganized as an independent International Astrostatistics Association, it has over 160 members from both the astronomical and statistical communities. They organized astrostatistics sessions at the 58th ISI Congress in August 2011 in Dublin (http://www.isi2011.ie/content/). At the Beijing General Assembly, the IAU formed a sister organization, the Working Group in Astrostatisitcs and Astroinformatics, to reside under Commission 5 in Division B.

Each session will have at least one keynote speaker, and half of the talks will be contributed. The cross-disciplinary nature of the symposium is reflected in the inclusion of speakers from the statistics community in the suggested list.

Special sessions:
a) Poster viewings and juried competition for best poster with prize.
b) Afternoon/evening tourism (Estoril coast & beaches or Lisbon visit, Fado dinner/concert).
c) Round table discussion of accomplishments and future of statistical cosmology. Special topics for discussion: Evolving best practices for dealing with petabyte-scale data such as (a) what data are made available (likelihoods, probabilistic catalogues), (b) open access, (c) reproduceability standards.