Anomaly Detection for Multivariate Time Series of Exotic Supernovae
This addresses the need for real-time anomaly detection in astronomy to flag unusual supernovae for followup study, though it is incremental as it combines existing methods like RNN-based variational autoencoders and isolation forests.
The authors tackled the problem of rapidly identifying astrophysically interesting supernovae from large datasets by developing an unsupervised anomaly detection pipeline for multivariate time series, successfully discovering anomalous supernovae and objects with incorrect redshift measurements in a simulated dataset of 12,159 supernovae.
Supernovae mark the explosive deaths of stars and enrich the cosmos with heavy elements. Future telescopes will discover thousands of new supernovae nightly, creating a need to flag astrophysically interesting events rapidly for followup study. Ideally, such an anomaly detection pipeline would be independent of our current knowledge and be sensitive to unexpected phenomena. Here we present an unsupervised method to search for anomalous time series in real time for transient, multivariate, and aperiodic signals. We use a RNN-based variational autoencoder to encode supernova time series and an isolation forest to search for anomalous events in the learned encoded space. We apply this method to a simulated dataset of 12,159 supernovae, successfully discovering anomalous supernovae and objects with catastrophically incorrect redshift measurements. This work is the first anomaly detection pipeline for supernovae which works with online datastreams.