A0978
Title: Astro-statistical learning and uncertainty quantification
Authors: Joshua Speagle - University of Toronto (Canada) [presenting]
Abstract: Astrophysics is entering a regime where the theoretical models of many astrophysical phenomena are noticeably less expressive than the rich and massive amounts of data we are collecting. As a result, there is increasing interest in leveraging statistical learning methods to simultaneously uncover underlying structure while performing latent parameter inference. The aim is to discuss some of the challenges in accomplishing these goals within astrophysical datasets, which span data characteristics (heterogeneous measurements, biased and censored sampling, etc.), uncertainty quantification (calibration, covariate shift, etc.), and model selection (ill-posed model classes, etc.), among others. Examples are motivated by particular astrophysical inference problems, including ongoing efforts towards practical solutions.
