Classification of multiwavelength transients with Machine Learning

With the advent of powerful telescopes such as the SKA and LSST, we are entering a golden era of multiwavelength transient astronomy. In order to cope with the dramatic increase in data volume as well as successfully prioritise spectroscopic follow-up resources, we propose a new machine learning approach for the classification of radio and multiwavelength transients. The algorithm consists of three steps: (1) augmentation and interpolation of the data using Gaussian processes; (2) feature extraction using a wavelet decomposition; (3) classification with the robust machine learning algorithm known as random forests. We apply this algorithm to existing radio transient data, illustrating its ability to accurately classify most of the eleven classes of radio variables and transients after just eight hours of observations, achieving an overall accuracy of $73.5\%$. We show how performance is expected to increase as more training data are acquired, by training the classifier on a simulated representative training set, achieving an overall accuracy of $97.4\%$. Finally, we outline a general approach for including multiwavelength data for general transient classification, and demonstrate its effectiveness by incorporating a single optical data point into the analysis, which improves the overall accuracy by $\approx 22\%$.