Transient-optimized real-bogus classification with Bayesian convolutional neural networks – sifting the GOTO candidate stream

Killestein, Thomas and Lyman, Joe and Steeghs, D and Ackley, K and Dyer, Martin J and Ulaczyk, K and Cutter, R. and Mong, Y-L and Galloway, D K and Dhillon, Vik and O’Brien, P and Ramsay, G and Poshyachinda, S and Kotak, R and Breton, Rene P and Nuttall, L K and Pallé, E and Pollacco, D and Thrane, E and Aukkaravittayapun, S and Awiphan, S and Burhanudin, U and Chote, P and Chrimes, A and Daw, E and Duffy, Christopher and Eyles-Ferris, R and Gompertz, Benjamin and Heikkilä, T and Irawati, P and Kennedy, Mark R and Levan, A and Littlefair, S and Makrygianni, L and Mata Sánchez, D and Mattila, S and Maund, Justyn and McCormac, J and Mkrtichian, D and Mullaney, J and Rol, E and Sawangwit, U and Stanway, Elizabeth and Starling, Rhaana and Strøm, P A and Tooke, S and Wiersema, K and Williams, Steven C. (2021) Transient-optimized real-bogus classification with Bayesian convolutional neural networks – sifting the GOTO candidate stream. Monthly Notices of the Royal Astronomical Society, 503 (4). pp. 4838-4854. ISSN 0035-8711

Abstract

Large-scale sky surveys have played a transformative role in our understanding of astrophysical transients, only made possible by increasingly powerful machine learning-based filtering to accurately sift through the vast quantities of incoming data generated. In this paper, we present a new real-bogus classifier based on a Bayesian convolutional neural network that provides nuanced, uncertainty-aware classification of transient candidates in difference imaging, and demonstrate its application to the datastream from the GOTO wide-field optical survey. Not only are candidates assigned a well-calibrated probability of being real, but also an associated confidence that can be used to prioritize human vetting efforts and inform future model optimization via active learning. To fully realize the potential of this architecture, we present a fully automated training set generation method which requires no human labelling, incorporating a novel data-driven augmentation method to significantly improve the recovery of faint and nuclear transient sources. We achieve competitive classification accuracy (FPR and FNR both below 1 per cent) compared against classifiers trained with fully human-labelled data sets, while being significantly quicker and less labour-intensive to build. This data-driven approach is uniquely scalable to the upcoming challenges and data needs of next-generation transient surveys. We make our data generation and model training codes available to the community.