AstroAI Lunch Talks - March 4, 2024 - Kaylee de Soto
04 Mar 2024 - Joshua Wing
The video can be found here: https://www.youtube.com/watch?v=2o1h_Oy1mpc
Speaker: Kaylee de Soto
Title: Exploring Machine Learning Pipelines for Realtime Supernova Classification
Abstract: Expected Rubin alert rates necessitate the development of reliable and efficient pipelines to categorize transients for spectroscopic follow-up. Here, we discuss some methods using machine learning to classify supernovae from their photometry. We mainly focus on Superphot+, which fits multiband light curves to an empirical model, and uses the resulting model parameters to train a gradient-boosted machine. Superphot+ takes advantage of nested sampling and stochastic variational inference to fit data from ZTF and Rubin datastreams in real time. We discuss performance of Superphot+ on archival ZTF data and future adaptation to six-band Rubin light curves. We also discuss current efforts to encode light curves using nonparametric techniques.
Watch the talk below!