Machine learning for transient recognition in difference imaging with minimum sampling effort

Mong, Y-L and Ackley, K and Galloway, D K and Killestein, T and Lyman, Joe and Steeghs, D and Dhillon, Vik and O’Brien, P T and Ramsay, G and Poshyachinda, S and Kotak, R and Nuttall, L and Pallé, E and Pollacco, D and Thrane, E and Dyer, Martin J and Ulaczyk, K and Cutter, R. and McCormac, J and Chote, P and Levan, A J and Marsh, T and Stanway, Elizabeth and Gompertz, Benjamin and Wiersema, K and Chrimes, Ashley and Obradovic, A and Mullaney, J and Daw, E and Littlefair, S and Maund, Justyn and Makrygianni, L and Burhanudin, U and Starling, Rhaana and Eyles-Ferris, R A J and Tooke, S and Duffy, C and Aukkaravittayapun, S and Sawangwit, U and Awiphan, S and Mkrtichian, D and Irawati, P and Mattila, S and Heikkilä, T and Breton, Rene P and Kennedy, Mark R and Mata Sánchez, D and Rol, E (2020) Machine learning for transient recognition in difference imaging with minimum sampling effort. Monthly Notices of the Royal Astronomical Society, 499 (4). pp. 6009-6017. ISSN 0035-8711

Abstract

ABSTRACT <jats:p/>The amount of observational data produced by time-domain astronomy is exponentially increasing. Human inspection alone is not an effective way to identify genuine transients from the data. An automatic real-bogus classifier is needed and machine learning techniques are commonly used to achieve this goal. Building a training set with a sufficiently large number of verified transients is challenging, due to the requirement of human verification. We present an approach for creating a training set by using all detections in the science images to be the sample of real detections and all detections in the difference images, which are generated by the process of difference imaging to detect transients, to be the samples of bogus detections. This strategy effectively minimizes the labour involved in the data labelling for supervised machine learning methods. We demonstrate the utility of the training set by using it to train several classifiers utilizing as the feature representation the normalized pixel values in 21 × 21 pixel stamps centred at the detection position, observed with the Gravitational-wave Optical Transient Observer (GOTO) prototype. The real-bogus classifier trained with this strategy can provide up to $95{{\ \rm per\ cent}}$ prediction accuracy on the real detections at a false alarm rate of 1%.