Vetting the optical transient candidates detected by the GWAC network using convolutional neural networks

Abstract

The observation of the transient sky through a multitude of astrophysical messengers has led to several scientific breakthroughs these last two decades thanks to the fast evolution of the observational techniques and strategies employed by the astronomers. Now, it requires to be able to coordinate multi-wavelength and multi-messenger follow-up campaigns with instruments both in space and on ground jointly capable of scanning a large fraction of the sky with a high imaging cadency and duty cycle. In the optical domain, the key challenge of the wide field of view telescopes covering tens to hundreds of square degrees is to deal with the detection, the identification and the classification of hundreds to thousands of optical transient (OT) candidates every night in a reasonable amount of time. In the last decade, new automated tools based on machine learning approaches have been developed to perform those tasks with a low computing time and a high classification efficiency. In this paper, we present an efficient classification method using Convolutional Neural Networks (CNN) to discard many common types of bogus falsely detected in astrophysical images in the optical domain. We designed this tool to improve the performances of the OT detection pipeline of the Ground Wide field Angle Cameras (GWAC) telescopes, a network of robotic telescopes aiming at monitoring the optical transient sky down to R=16 with a 15 seconds imaging cadency. We applied our trained CNN classifier on a sample of 1472 GWAC OT candidates detected by the real-time detection pipeline.

O’TRAIN: A robust and flexible ‘real or bogus’ classifier for the study of the optical transient sky

“…These probabilities can then be easily interpreted by astronomers to trigger further targeted follow-up observations. Usually, in time-domain astronomy, ML algorithms are used to perform the two following classification tasks: (1) 'real or bogus' classification to reduce the false alarm rate due to artefacts that may be falsely identified as new real sources (Gieseke et al 2017;Duev et al 2019;Turpin et al 2020;Killestein et al 2021;Hosenie et al 2021) and (2) astrophysical classification to identify the types of transients the detection pipelines have detected based on several pieces of information about some key parameters evolving with time, the flux-colour evolution of the transient or its spectral shape and associated features, etc. (Carrasco-Davis et al 2019, 2021Möller & de Boissière 2020;Burhanudin et al 2021).…”

Alert Classification for the ALeRCE Broker System: The Real-time Stamp Classifier

“…To the best of our knowledge, this is the first classifier that discriminates among five classes using a single alert, allowing a rapid, reliable characterization of the data stream to trigger immediate follow-up. Previous work on stamp classification has focused instead on the classification of real objects vs. bogus detections (e.g., Goldstein et al 2015;Cabrera-Vives et al 2017;Reyes et al 2018;Duev et al 2019;Turpin et al 2020), galaxy morphologies (e.g., Dieleman et al 2015;Prez-Carrasco et al 2019;Barchi et al 2020), or time domain classification (Carrasco-Davis et al 2019;Gmez et al 2020).…”

O'TRAIN: a robust and flexible Real/Bogus classifier for the study of the optical transient sky

“…1. Real / bogus classification to reduce the false alarm rate due to artefacts that may be falsely identified as new real sources (Gieseke et al 2017;Duev et al 2019;Turpin et al 2020;Killestein et al 2021;Hosenie et al 2021) 2. Astrophysical classification to identify the types of transients the detection pipelines have detected based on several pieces of information about some key parameters evolving with time, the flux-color evolution of the transient or its spectral shape and associated features, etc.…”

“…In this paper, we propose a robust machine learning algorithm called O'TRAIN for Optical TRAnsient Identification Network, to filter out any type of bogus from a list of optical transient (OT) candidates a detection pipeline may output. Our real/bogus (RB) classifier is based on a Convolutional Neural Network (CNN) algorithm, a method that already proved its efficiency in such a classification task (Gieseke et al 2017;Turpin et al 2020). We developed a lot of pedagogical tools to easily launch a training procedure and diagnose the classification performances.…”