![]() In the early stages of these surveys candidate identification was performed manually, with humans ‘blinking’ images to look for varying sources. Many of the earliest prototypical transient surveys began as galaxy-targeted searches, performed with small field-of-view instruments. Requiring multiple observations of the same sky area to detect variability, transient surveys naturally generate vast quantities of data that require processing, filtering, and classification – this has driven the development of increasingly powerful techniques bolstered by machine learning to meet the demands of these projects. 1999), and located the faint counterparts of some of the most distant and energetic astrophysical events known: gamma-ray bursts (e.g. 2017), traced the accelerating expansion of our Universe across cosmic time (e.g. Through observations of optical transient sources we have obtained evidence of the explosive origins of heavy elements (e.g. This variability can occur on time-scales of milliseconds to years, and at luminosities ranging from stellar flares to luminous supernovae that outshine their host galaxy (Kulkarni 2012 Villar et al. Transient astronomy seeks to identify new or variable objects in the night sky, and characterize them to learn about the underlying mechanisms that power them and govern their evolution. Methods: data analysis, techniques: photometric, surveys 1 INTRODUCTION We make our data generation and model training codes available to the community. This data-driven approach is uniquely scalable to the upcoming challenges and data needs of next-generation transient surveys. 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. 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. 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. ![]() 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. 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.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |