Previous | Session 29 | Next | Author Index | Block Schedule
E. Devinney, E. Guinan (Villanova University), D. Bradstreet (Eastern University), M. DeGeorge (Panasonic Corp.), J. Giammarco (Temple University), C. Alcock (Harvard University), S. Engle (Villanova University)
The explosive growth of observational capabilities and information technology over the past decade has brought astronomy to a tipping point - we are going to be deluged by a virtual fire hose (more like Niagara Falls!) of data. An important component of this deluge will be newly discovered eclipsing binary stars (EBs) and other valuable variable stars. As exploration of the Local Group Galaxies grows via current and new ground-based and satellite programs, the number of EBs is expected to grow explosively from some 10,000 today to 8 million as GAIA comes online. These observational advances will present a unique opportunity to study the properties of EBs formed in galaxies with vastly different dynamical, star formation, and chemical histories than our home Galaxy. Thus the study of these binaries (e.g., from light curve analyses) is expected to provide clues about the star formation rates and dynamics of their host galaxies as well as the possible effects of varying chemical abundance on stellar evolution and structure. Additionally, minimal-assumption-based distances to Local Group objects (and possibly 3-D mapping within these objects) shall be returned. These huge datasets of binary stars will provide tests of current theories (or suggest new theories) regarding binary star formation and evolution. However, these enormous data will far exceed the capabilities of analysis via human examination.
To meet the daunting challenge of successfully mining this vast potential of EBs and variable stars for astrophysical results with minimum human intervention, we are developing new data processing techniques and methodologies. Faced with an overwhelming volume of data, our goal is to integrate technologies of Machine Learning and Pattern Processing (Artificial Intelligence [AI]) into the data processing pipelines of the major current and future ground- and space-based observational programs. Data pipelines of the future will have to carry us from observations to astrophysics with minimal human intervention. While there has been some recognition of this need (e.g. the LSST project drawing on the experience of MACHO/OGLE), few steps have been taken to address this crucial issue.
Fortunately, advances in AI have created the opportunity to make significant progress in this direction. Here we discuss our plans to develop an Intelligent Data Pipeline (IDP) that can operate autonomously on large observational datasets to produce results of astrophysical value. Plans and initial results are discussed. This research is supported by NSF/RUI Grant AST05-07542 which we gratefully acknowledge.
Previous | Session 29 | Next
Bulletin of the American Astronomical Society, 37 #4
© 2005. The American Astronomical Soceity.