This is a guest post by Matthew Turk, an NSF Postdoctoral Fellow at Columbia University.
This summer saw the twelfth edition of the annual SciPy conference in Austin, Texas from June 24th to 29th. Python users will recognize the name of the SciPy package, which includes a core scientific toolkit, but the SciPy conference has a much broader scope: to bring together the whole community of developers of open source Python software for mathematics, science, and engineering; to showcase the latest projects; and to collaborate on code development. The conference program typically consists of tutorials (ranging from introductory to advanced levels), keynote talks, parallel sessions, mini-symposia on specific topics, and coding sprints where participants are encouraged to contribute to open source projects.
All of the talks are listed here and also posted to YouTube. Links to the YouTube videos can also be found on the individual talk pages (e.g. Astropy and SunPy) which are accessible from the full program. Many more conference ephemera (e.g., photos, tweets) can be found on eventifer.
This year, Tom Aldcroft (CfA) ran a great mini-symposium about astronomy. These talks covered many aspects of both astronomy and Python including: using Python for the LSST and SKA (and the gritty technical details of their implementation stacks!), querying the Virtual Observatory with PyVO, presentations about Astropy and SunPy (core libraries for Astronomy and Solar Physics respectively), and the image viewer Ginga. In addition to the mini-symposium talks, a few Astronomy-specific talks were given in the plenary session:
- “Multidimensional Data Exploration with Glue” by Chris Beaumont
- “Parallel Volume Rendering in yt: User Driven & User Developed” by Sam Skillman
- “Opening Up Astronomy with Python and AstroML” by Jake Vanderplas
- “Bringing Astronomical Tools down to Earth” by Mike Droettboom
For anyone who is not familiar with these packages, Glue is a Python library that allows users to explore relationships within and across datasets, yt is a powerful package that enables detailed data analysis and visualization of simulations, and AstroML is a package for machine learning and data mining.
While there were many other great talks about a diverse range of technical and scientific topics, a few really stood out during the course of the week:
- “Why you should write buggy software with as few features as possible” by Brian Granger
- “Writing Reproducible Papers with Dexy” by Ana Nelson
Last year, James Turner organized the astro mini-symposium, which also had a number of great talks, including (shameless self-promotion!) mine and Perry Greenfields‘, along with lots of other good ones, scattered throughout this page.
In my experience, one of the biggest challenges in astronomy–where science projects seem to be segregated into either completely vertically-integrated stacks or enormous projects–is to figure out the best ways to build collaborations and leverage opportunities. Outreach can be a difficult thing; perhaps it’s different in observation, where the “secret sauce” is slightly more removed than in simulation fields, but on our end figuring out how to engage people working on similar projects and perhaps even in direct competition can be difficult, although certainly surmountable. I personally believe that, in simulations, community is the next greatest challenge — not because it will help scale to “exascale” processors or manage “big data” or develop new cyberinfrastructure, but because the long tail of computational science is where the vast majority of scientific discovery occurs and now, more than ever, community and collaboration is essential to ensuring the health of that ecosystem.
The Scientific Python community is an amazing example of that and the SciPy conference is an excellent way to develop those collaborations within and between domains. The strongest message–perhaps more than “Python is winning”–was “open source and collaboration are winning.” Furthermore, the SciPy conference has also been actively attempting to grow the diversity of its participants. Despite being, in my experience, somewhat uncommon in most computational science conferences, this is essential to the health of a community.