LECTURES A GRATIS GLOBAL SERVICE
⌕ SEARCH GRATIS GLOBAL ↗
LECTURES
Generative Solutions for Cosmic Problems
SOURCE: YOUTUBE · NO TRACKING UNTIL YOU PRESS PLAY · TROUBLE PLAYING? WATCH AT THE SOURCE ↗

Generative Solutions for Cosmic Problems

60 MIN · EN · STATUS: [ STREAMING ]
RATE THIS
IAS

Carolina Cuesta-Lazaro, speaking in the Institute for Advanced Study Astrophysics Seminar, surveys her research on generative machine learning methods in cosmology. She covers how generative models enable high-dimensional simulation-based inference, building toward field-level inference, and how machine-learned emulators of expensive hydrodynamical simulations unlock large-scale analysis of observables such as the Sunyaev-Zel'dovich effects, fast radio bursts, and weak lensing baryonification. She discusses techniques for learning low-dimensional representations that separate shared physical signal from instrument-specific noise in a joint latent space, aimed at more robust inference between simulations and real observations. Time permitting, she closes with a look at how LLM agents might automate parts of the scientific method by proposing and testing cosmological theories. The talk is aimed at an audience familiar with cosmological simulation and inference methods, moving quickly through technical material rather than offering a general introduction.

At a glance

Lecture facts

Runtime compared with the other 387 Computer Science lectures
Runtime1 h
Compared with Computer ScienceShorter than 66%
Source channelInstitute for Advanced Study (YouTube)