|
|
|
Facilitated Access and Application of Computational Astrostatistics Algorithms
|
|
|
| Astronomy and Astrophysics are experiencing a revolution with the combination of huge data resources and the drive for high
precision measurements. A part of this revolution is new algorithms for the analysis of astronomical data which scale--up into the
regime of billions of sources with thousands of dimensions. To address this need, we have established an interdisciplinary research
team of astrophysicists, statisticians and computer scientists: the INternational Computational Astrostatistics (INCA) group. This
group is responsible for developing fast and efficient statistical algorithms for astrophysics. Key to the success of INCA is the
dissemination of these new algorithms to the research communities. This dissemination requires much more than making available a
tar-ball package of code and some README files. The algorithms and techniques being built to handle modern and future
astronomical datasets, are often complex in both design and implementation. Therefore, we propose here to build a set of
well-documented libraries for the most commonly used scripting languages available to astronomers and statisticians. This includes
exporting the INCA algorithms (which include KD-tree codes, fast mixture model algorithms and non--parametric fitting routines) to
IDL, R and Python. In particular, IDL has gained phenomenal popularity within the astronomical community over the last 10 years.
The goal is to provide a majority of astronomical researchers the opportunity to trivially and quickly apply these new techniques to
their data and rapidly visualize the results. Just as importantly, we will export common astronomical algorithms to the statistical R
scripting-language environment. In doing so, we can facilitate new advances from statisticians which will flow back quickly to the
astronomical community, as they can be developed against astronomical challenges. Finally, and perhaps most importantly, we will
deploy these algorithms as service-based web services (i.e., in a Service Oriented Architecture or SOA). This means that the analysis
tools can be placed where the large data volumes reside. This is a paradigm shift in astronomy where we "move the services to the
data", as opposed to moving the data to our desktops. Along with the web service protocols, we will provide browser-based user
interface mechanisms for utilizing these amazing new algorithms. |
|
|
