|
|
| The EBAI Project: Testing Artificial Intelligence / Neural Network Approaches to Automatically Solve Light Curves of Eclipsing Binary Systems |
Edward F. Guinan1, N. Hollon1, A. Prsa1, E. Devinney1
1Villanova Univ.. |
| Presentation Number: 003.36 |
Advances in observing technology will greatly increase discovery rates of eclipsing binaries (EBs). For example, missions such as LSST and GAIA are expected to yield hundreds of thousands (even millions) of new EBs. Current personal interactive (and time consuming) methods of determining the orbital and physical parameters of EBs from their light curves will be totally inadequate to keep up with the overwhelming flood of new data. At present the currently used methods require significant technical skill, and even experienced light curve solvers take 2-3 weeks to model a single binary. We are therefore developing an Artificial Intelligence / Neural Network system with the hope of creating a fully automated, high throughput process for gleaning the orbital and physical properties of EB-systems from the observations of tens of thousands of eclipsing binaries at a time. This project is called EBAI - Studying Eclipsing Binaries with Artificial Intelligence (See: http://www.eclipsingbinaries.org). A preliminary test of the neural network's performance has been conducted, using as input the normalized Johnson V-filter flux curves for five detached EBs: KP Aql, AY Cam, WX Cep, DI Her, and BP Vul. These systems have well determined properties from previous detailed photometric and radial velocity analyses. The neural network system has met with promising success in analyzing these systems. The results of this test and additional tests on larger samples of stars will be presented and discussed.
This research is supported by NSF/RUI Grant No. AST-05-07542 which we gratefully acknowledge. |
| |
|
|
|
|
|
|
The Online Abstract Submission and Invitation System
© 1996 - 2010 Coe-Truman Technologies, Inc. All rights reserved.
|
|
|
|
|
|
