CCA Colloquium: Uros Seljak

Title: Optimal knowledge extraction in the era of massive astronomical datasets

Abstract: Astronomy is experiencing a renaissance driven by extraordinary data volumes flowing from both terrestrial and space-based surveys. In parallel, our computational frameworks have evolved in complexity and capability. A further opportunity for advancing our scientific understanding lies in extracting information from these surveys with optimal precision and reliability. I will showcase multiple approaches to this challenge, spanning from sophisticated statistical methodologies—including high-dimensional optimization and sampling techniques—to cutting-edge AI and machine learning applications. I will present applications to cosmology, particularly in weak lensing and galaxy clustering analyses, as well as to exoplanet detections. The results suggest these advanced methods substantially outperform traditional approaches. Methodologies developed for astronomical applications often prove valuable beyond their original context. An example of such interdisciplinary transfer is our recently developed MicroCanonical Langevin and Hamiltonian Monte Carlo, which is finding applications across multiple scientific domains.