EBFactory/RRLFactory is an extendable end to end pipelane for period finding, classification and post-processing stellar light curves. We are developping a tool allowing to upload light-curves, searching for periodic signals and classifying signals found by several neural networks. A more detailed description is here.
We currently concentrate on Eclipsing Binaries and RR-Lyrae type variables. The panel graphics show sample light curves for Eclipsing binaries (top left: contact, top right: detached) and RR Lyrae (bottom left: RRc, bottom right RRab).
Light curves identified as binaries are post-processed by a “Solution Estimator”, another neural network developped and trained by Andrej Prsa at Villanova University, yielding an estimate for properties such as relative temperatures and sizes.
This pipeline is able to classify and estimate massive amounts of light curves as produced (several thousands per second on a desktop class PC) and to sort out the “interesting” systems such as an RR-Lyra type variable as part of an Eclipsing binary.
The histogram shows the retrieval rates and false positives for several main categories of variable stars.
The goal of EBFactory is to manage masses of data and light curves as produced by Kepler and the oncoming LSST.