Image-based Classification of Variable Stars: First Results from Optical Gravitational Lensing Experiment Data

Recently, machine learning methods have presented a viable solution for the automated classification of image-based data in various research fields and business applications. Scientists require a fast and reliable solution in order to handle increasingly large amounts of astronomical data. However, so far astronomers have been mainly classifying variable starlight curves based on various pre-computed statistics and light curve parameters. In this work we use an image-based Convolutional Neural Network to classify the different types of variable stars. We use images of phase-folded light curves from the Optical Gravitational Lensing Experiment (OGLE)-III survey for training, validating, and testing, and use OGLE-IV survey as an independent data set for testing. After the training phase, our neural network was able to classify the different types between 80% and 99%, and 77%-98%, accuracy for OGLE-III and OGLE-IV, respectively.