|
|
|
Knowledge Discovery in a Virtual Universe
|
|
|
| Astrophysics is witnessing a flood of data from new ground and space based telescopes and surveys. With access to data sets spanning X-rays through to radio wavelengths we will soon be able to extract sources from any region of the sky and measure their properties across the full range of the electromagnetic spectrum. Individually, each of these data sets has the potential to advance our understanding of the processes that drive the formation and evolution of our Universe. It is, however, only when these data are combined that their full scientific potential will finally be realized; the scientific returns from the total will far exceed those from any one individual component. The recognition of this by the Astronomical community has led to a major new initiative: the National Virtual Observatory (NVO; http//www.us-vo.org). If we are to realize the full potential of the era of the NVO we must be able to explore, analyze and interact with these massive datasets. We need tools for knowledge discovery that not only make data delivery fast and easy to learn but also enable rapid data analysis on a massive scale. The resources required to analyze these terabyte and larger databases can be found if we can harness the power of the upcoming generation of distributed computing resources (i.e. the Grid). The challenge we face is how do we help a user to realize the scientific value of the data without being limited by the laborious and complicated efforts required to implement data analysis techniques that span many processors distributed across a grid? Can we not only make data delivery fast and easy to learn (as the NVO is doing), but can we also minimize the time that it takes an astronomer to use that data to actually find an answer? This proposal will address the challenge of knowledge discovery in the era of the NVO and the Grid; providing the final component necessary to link the massive astrophysical data sources to massive distributed computational resources. We will develop new parametric and non-parametric algorithms and techniques that solve common problems in astrophysical data analysis and that are optimized for distributed computing environments. This research program has the potential to impact not only space science and the NVO, but also any NASA missions that generate large multidimensional datasets. |
|
|
