A machine learning approach for GRB detection in AstroSat CZTI data

We present a machine learning (ML) based method for automated detection of Gamma-Ray Burst (GRB) candidate events in the range 60 KeV - 250 KeV from the AstroSat Cadmium Zinc Telluride Imager data. We make use of density-based spatial clustering to detect excess power and carry out an unsupervised hierarchical clustering across all such events to identify the different categories of light curves present in the data. This representation helps in understanding the sensitivity of the instrument to the various GRB populations and identifies the major non-astrophysical noise artefacts present in the data. We make use of dynamic time wrapping (DTW) to carry out template matching to ensure the morphological similarity of the detected events with that of known typical GRB light curves. DTW alleviates the need for a dense template repository often required in matched filtering like searches, and the use of a similarity metric facilitates outlier detection suitable for capturing un-modeled events previously. The developed pipeline is used for searching GRBs in AstroSat data, and we briefly report the characteristics of newly detected 35 long GRB candidates. With minor modifications such as adaptive binning, the method is also sensitive to short GRBs events. Augmenting the existing on-ground data analysis pipeline with such ML capabilities could alleviate the need for extensive manual inspection and thus enable quicker response to alerts received from other observatories such as the gravitational-wave detectors.