ASAIP Editors' blog, October 2013
Greetings! This month's blog looks at meetings in astrostatistics & astroinformatics, both recently past and future meetings. We have reports by Eric Ford on the SAMSI workshop on the Kepler studies of exoplanets, and by Joe Hilbe on the astrostatistics sessions at the recent World Statistics Congress. And we have a followup to the last blog's discussion of image deconvolution.
Modern Statistical and Computational Methods for Analysis of Kepler Data, June 10-28 2013, SAMSI, Research Triangle NC USA
For those unfamiliar, SAMSI is NSF's Statistics and Applied Mathematics Science Institute associated with three universities: Duke, UNC and NCSU. It is an empty building that periodically fills with cross-disciplinary experts to address forefront problems in many fields: economics, homeland security, genomics, environment … and astrophysics. Workshops last weeks or even months with resident researchers.
NASA’s Kepler Mission was designed to search for small planets, including those in the Habitable Zone of sun-like stars. The search was conducted with a specially designed telescope situated aboard a space craft on a heliocentric orbit. From 2009-2013, the telescope observed over 190,000 stars nearly continuously, reporting their brightness once every 1 or 30 minutes with extraordinarily high-precision. This photometry is revolutionizing several areas of astronomy, including exoplanets when they pass in front of the star, astroseismology using rapid pulsations of stars, and other types of variable stars. It is also raising several new statistical challenges which were addressed at the SAMSI Kepler Program. Statisticians learnt about astronomy and the ﬁne points of the Kepler data, while astronomers learnt about advanced statistical techniques. Then members from both groups started immediate collaborations to put statistics to good use in the search for small, rocky and potentially habitable planets.
Two small working groups focused on planet characterization and long-period planets and two larger groups. Two larger working groups are working on characterizing exoplanet populations with Bayesian hierarchical models, developing and testing various algorithms for detrending of light curves to improve planet detection. Some of the methodology is already quite sophisticated, while other aspects are just emerging.
Astrostatistics at the 59th World Statistics Congress, 25-30 August 2013, Hong Kong CN
The ISI is the International Statistical Institute, the oldest and biggest worldwide statistics association. It plays the same role in statistics as the IAU in astronomy. The ISI has a biannual convention called The World Statistics Congress (WSC), which was held this year in Hong Kong; the 2015 WSC will be in Rio de Janerio.
Astrostatistics was awarded an invited papers and a special topics session at the past WSC. The title of the invited papers session, organized by Bodhisattvai Sen of Columbia University was, "Statistical challenges in astronomy". The session was chaired by Daniel Mortlock (Imperial College, London), and given on 28 August. Invited papers were delivered by Jogesh Babu (Pennsylvania State University) on Big Data in astronomy, Thomas C.M. Lee (University of California, Davis) on semi-parametric estimation of nonstationarity in high energy time series, Debashis Mondal (University of Chicago) on wavelet analysis of irregularly sampled time series, and Joseph Hilbe (Arizona State University) on the impact of astrostatistics on statistics.
The Special Topics session was organized by Hilbe and chaired by Xiaodan Fan (Chinese University of Hong Kong). The session, presented on 27 August, was titled, "Source and feature detection in astronomy". Papers were delivered by Roberto Trotta (Imperial College, London) on Dark Matter signals in Fermi gamma-ray data, Daniel Mortlock on Bayesian hypothesis test without a previous alternative model, and Byron Bell (DePaul University) on classification of quasars. Trotta also presented a paper prepared by Farhan Feroz and Michael Hobson (University of Cambridge), who unavoidably could not be present, on Bayesian object detection in noisy data.
Members of the International Astrostatistics Association (IAA) gathered at the World Statistics Congress. Items such as forming committees of interest group were discussed. Hilbe announced that a corporate grant of $10,000 was awarded to the IAA for 2014, and additional funding is being sought.
Some forthcoming meetings that might interest you ...
Hot-wiring the Transient Universe III (13-15 November 2013, Sante Fe NM USA) will explore massively parallel time domain astronomical surveys coupled with rapid coordinated multi-wavelength follow-up observations.
Astroinformatics 2013: Knowledge from Data (9-13 December 2013, Sydney AU) will focus on advanced machine-learning and data-mining techniques for the new generation radio telescopes ASKAP and MeerKAT.
OBayes: Celebrating 250 Years of Bayes (15-17 December 2013, Durham NC USA) is a multi-faceted celebration of Bayes Theorem with an emphasis on objective Bayes techniques.
MCMski IV (6-9 January 2014, Chamonix FR) concentrates on Monte Carlo Markov Chain methodology including a session on astrophysical applications and a day on Bayesian nonparametrics.
2014 SIAM International Conference on Data Mining (24-26 April, Philadelphia PA USA) is a large cross-disciplinary meeting.
Statistical Challenges in 21st Century Cosmology: IAU Symposium 306 (25-29 May 2014, Lisbon PT) is the first IAU Symposium on astrostatistics, sponsored by the IAU Working Group in Astrostatistics and Astroinformatics. Topics include the cosmic microwave background, weak lensing, large-scale structure, high-redshift supernovae, Lyman-alpha forest, and other large datasets.
A followup to the August blog ….
Following our discussion on Bayesian improvements to EM-type algorithms for image reconstruction (known in astronomy as the Lucy-Richardson algorithm), Mario Bertero of the University of Genoa shared with us some closely related papers by the Italian Research Group on Optimization Algorithms and Software for Inverse Problems. Here several constraints are made on the point spread function: non-negativity, normalization and a bound based on the Strehl ratio. The objective function is the Kullback-Leibler divergence and computation uses a scaled gradient projection method. See their papers here and here.