Astronomy Department Calendar

Transients exist at the intersection of a wide range of disciplines in physics and astrophysics.  From standard candles providing constraints on cosmic expansion to explosive nucleosynthesis teaching us about the origins of the elements, the study of transients has led to discoveries at all scales.   In the age of the Vera C. Rubin Observatory the number of transient discoveries will explode by orders of magnitude.  Therefore, we require tools that can rapidly process millions of transient alerts per night to identify, classify, and forecast the evolution of transients for spectroscopic follow-up.  Supernova Explosions Learned by Deep ODE Networks (SELDON) is a transient foundation model built from the ground-up specifically for these tasks.  SELDON consists of an RNN-ODE encoder, which performs the difficult task of compressing variable length, irregularly sampled, multi-channel light curves into fixed-length vector embeddings suitable for a variety of downstream tasks.  SELDON's forecaster provides predictions for the evolution of the light curve at all times, past and future, along with estimates of physical parameters of the system in a self-consistent manner.  New extensions are being incorporated into SELDON to provide SED estimation, classification, and uncertainty quantification from light curve latent embeddings alone which will ready SELDON for real-time processing of Rubin alert streams.