Skip to main content
SearchLogin or Signup

Deep Learning for Multimessenger Astrophysics

Presentation #203.01 in the session “Machine Learning in Astronomy: Transient Discovery with Machine Learning (Meeting-in-a-Meeting)”.

Published onJun 18, 2021
Deep Learning for Multimessenger Astrophysics

The Vera C. Rubin Observatory’s (VRO) Legacy Survey of Space and Time (LSST) will discover hundreds of new transient events every night. The electromagnetic counterparts of the explosions resulting from the mergers of binary neutron star and neutron star-black hole systems, and other transients that have never been seen before lurk within this “alert stream.” Human inspection of the enormous volume of data is impossible, and spectroscopic observations for classification of all transients is already impossible with surveys that discover <1% of the transients that LSST will. These rare events also evolve much faster than LSST’s typical cadence, meaning that astronomers will need to identify these events for detailed follow-up from sparse data. I will describe how we are combining real data from existing surveys such as the Young Supernova Experiment and the Zwicky Transient Facility, the ANTARES alert broker system, and detailed realistic simulations of the time-domain sky as part of the Photometric LSST Astronomical Time-series Classification Challenge (PLAsTiCC) to train deep learning models capable of identifying these rarest of the rare events in real-time. I will also discuss how we can evolve this nascent cyberinfrastructure into a research platform for the analysis of observations from wide-field optical surveys, astroparticle experiments, and gravitational wave observatories – a development that will revolutionize and democratize how we do astrophysics.


No comments here