Computer Science – Learning
Scientific paper
Jan 2011
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2011aas...21734202d&link_type=abstract
American Astronomical Society, AAS Meeting #217, #342.02; Bulletin of the American Astronomical Society, Vol. 43, 2011
Computer Science
Learning
Scientific paper
The Eclipsing Binaries (EBs) via Artificial Intelligence (EBAI) Project is applying machine learning techniques to elucidate the nature of EBs. Previously, Prsa, et al. applied artificial neural networks (ANNs) trained on physically-realistic Wilson-Devinney models to solve the light curves of the 1882 detached EBs in the LMC discovered by the OGLE II Project (Wyrzykowski, et al.) fully automatically, bypassing the need for manually-derived starting solutions. A curious result is the non-monotonic distribution of the temperature ratio parameter T2/T1, featuring a subsidiary peak noted previously by Mazeh, et al. in an independent analysis using the EBOP EB solution code (Tamuz, et al.). To explore this and to gain a fuller understanding of the multivariate EBAI LMC observational plus solutions data, we have employed automatic clustering and advanced visualization (CAV) techniques.
Clustering the OGLE II data aggregates objects that are similar with respect to many parameter dimensions. Measures of similarity for example, could include the multidimensional Euclidean Distance between data objects, although other measures may be appropriate. Applying clustering, we find good evidence that the T2/T1 subsidiary peak is due to evolved binaries, in support of Mazeh et al.'s speculation. Further, clustering suggests that the LMC detached EBs occupying the main sequence region belong to two distinct classes. Also identified as a separate cluster in the multivariate data are stars having a Period-I band relation. Derekas et al. had previously found a Period-K band relation for LMC EBs discovered by the MACHO Project (Alcock, et al.). We suggest such CAV techniques will prove increasingly useful for understanding the large, multivariate datasets increasingly being produced in astronomy.
We are grateful for the support of this research from NSF/RUI Grant AST-05-75042 f.
DeGeorge M.
Devinney Edward J. Jr.
Guinan Edward F.
Prsa Andrej
No associations
LandOfFree
OGLE II Eclipsing Binaries In The LMC: Analysis With Class does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with OGLE II Eclipsing Binaries In The LMC: Analysis With Class, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and OGLE II Eclipsing Binaries In The LMC: Analysis With Class will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1402877