Last week, I had the great pleasure of attending the .Astronomy 4 meeting in Heidelberg, Germany. This was the first time I have attended this annual meeting, and I had the honor delivering one of the keynote addresses, on “How Can We Use Cloud Computing in Astronomy.” The goal of the .Astronomy conference series is “… to bring together an international community of astronomy researchers, developers, educators and communicators to showcase and build upon these many web-based projects, from outreach and education to research tools and data analysis.” Rather than have a formal agenda, .Astronomy mixes scheduled talks with “unconference” sessions on topics proposed at the meeting and with a “hack” day, devoted to small teams realizing new ideas.
Here I will give a summary of some of the new ideas – well, new to me that is – that I learned about at the meeting, rather than try to give a thorough report on the many events. There are other fine blog posts on the meeting, and I recommend reading them: .Astronomy 4 in der Haus, by Stuart Lowe; .Astronomy 4: TimTam Slams, Visual Goodness, and an Awful Lot of Code, by Sarah Kendrew; .Astronomy 4 in Heidelberg by Robert Simpson, and David Hogg provided summaries for Days One, Two and Three on his research blog.
On to the topics that took my eye. Kevin Govander talked about the big cultural impact of astronomy, and how the Square Kilometer Array is being used as a driver to get Africa connected. He talked about the work of the IAU Office of Astronomy for Development, which aims to implement an eight-year strategic plan for astronomy in the developing world. The OAD has already carried out a number of pilot projects: the VO demonstration Kevin mentioned took my eye.
Tom Robitaille talked about Astropy, a new project that is Python library for the astronomy community. It aims to to provide a core package of Python codes for astronomy. Packages such as PyFITS, PyWCS, vo, and asciitable already merged in. The source code is hosted through github.
Tom Kitching spoke of the power of using crowd sourcing to do science. The idea is simple: use collective intelligence to solve a difficult problem. He described how companies such as Kaggle and Crowd Flower have been set up to facilitate this type of enterprise. Kaggle’s approach is to set up competitions to solve problems – groups of participants can try different algorithms and see how they all perform. Tom described one such example: how weak lensing affects the shapes of galaxies, a difficult problem that has defeated teams of professional astronomers. The effect is small (1% or so), statistical in nature, and is of the same size as other effects such as distortions introduced by the telescope. Enter Kaggle, which set up a competition to measure the ellipticities of 60,000 galaxies, with a trip to JPL and authorship on a paper as the prize. The competition attracted 700 submissions fr0m 1o7 players, including particle physicists and Arabic handwriting experts. Techniques included neural nets and PCA analysis. The result was a x10 increase in accuracy in 2 months of competition.
A somewhat different approach to crowdsourcing was described by Michelle Borkin. The Amazon Mechanical Turk pays participants to take part. She cited the example of using crowdsourcing to analyze tree diagrams. Try it out at http://bit.ly/mturk_dotastronomy.