To date, three independent studies have devised an automatic procedure to analyse and extract the principal parameters of 2581 detached eclipsing binary stars from the OGLE photometric survey of the Large Magellanic Cloud (LMC): Devor (2005), Tamuz et al. (2006), and Prsa et al. (2008). For time efficiency, Devor used a simple model of two spherical, limb-darkened stars without tidal or reflection physics. Tamuz et al.'s approach employs a more realistic EBOP model, which is still limited in handling proximity physics. Our study used a back-propagating neural network that was trained on the light curves computed by a modern Wilson-Devinney code. The three approaches are confronted and correlations in the results are sought that indicate the degree of reliability of the obtained results. A database of solutions consistent across all three studies is presented. We assess the suitability of each method for other morphology types (i.e. semi-detached and overcontact binaries) and we overview the practical limitations of these methods for the upcoming survey data.
This research is supported by NFS/RUI Grant No. AST-05-07542, which we gratefully acknowledge.