Unsupervised Classification of Variable Stars

Abstract

During the last ten years, a considerable amount of effort has been made to develop algorithms for the automaticclassification of variable stars. That has been primarily achieved by applying machine learning methods to photometric datasets where objects are represented as light curves.Machine learning classifiers require training sets to learn the underlying patterns that allow the separation among classes.Unfortunately, building training sets for light curves is an expensive process that demands a lot of human efforts.Every time data comes from new surveys; the only available training instances are the ones that have a cross-match with previously labelled objects, consequently generating insufficient training sets compared with the large amounts of unlabelled sources. Also, many times astronomers pursue the identification of a particular type of variable stars, where is not reasonable to wait for training sets that are formed mostly by other classes of variability. In this work, we present an algorithm that performs unsupervised classification of variable stars, relying only on the similarity among light curves. We tackle the unsupervised classification problem by proposing an untraditional approach. Instead of trying to match classes of stars with clusters found by a clustering algorithm, we propose a query based method where astronomers can find groups of variable stars ranked by similarity. We also develop a fast similarity function specific for light curves, based on a novel data structure that allows scaling the search over the entire dataset of unlabelled objects. Experiments show that our unsupervised model achieves high accuracy in the classification of different types of variability and that the proposed algorithm scales up to massive amounts of light curves.