Back to Volume
Paper: Anomaly Detection in Catalogs of Periodic Variable Stars
Volume: 411, Astronomical Data Analysis Software and Systems XVIII
Page: 264
Authors: Rebbapragada, U.; Protopapas, P.; Brodley, C.E.; Alcock, C.
Abstract: We present an anomaly detection method for identifying potentially novel astronomical phenomena from catalogs of periodic variable stars. The challenge of working with large catalogs of light-curve data is that the data are unsynchronized, a fact which precludes the use of out-of-the-box anomaly detection methods. We present PCAD, a novel method whose primary innovation is the efficient integration of phase shift calculations with model calculation. Because our model calculation is robust on samplings of the data, PCAD scales to large data sets. An analysis of our results shows that PCAD is useful for identifying misclassifications in the catalogs, as well as truly unusual phenomena worthy of further investigation.
Back to Volume