Description
I will review the recent significant advances in statistical learning that have led to gravitational lensing analyses using "simulation-based inference" (SBI). These new statistical approaches permit data analyses that would otherwise be impossible. I will review results from the Kilo-Degree Survey and the Dark Energy Surveys, both using SBI for classical 2pt statistics and also beyond 2pt statistics (including AI-powered map-level results) – these have led to the tightest constraints (to-date) on dark energy from weak lensing data. I will cover the opportunities and challenges for these methods for Rubin LSST & Euclid.
[e.g. https://arxiv.org/abs/2212.04521, https://arxiv.org/abs/2404.15402, https://arxiv.org/abs/2403.02314 ]
Primary author
Niall Jeffrey
