Cosmo-26

In Search of New Physics in the CMB Using AI

Not scheduled

20m

Gorlaeus gebouw (Leiden University)

Talk Methods / Statistical Inference / Machine Learning

Speaker

Andrius Tamosiunas (University of Oslo)

Description

In this talk, I will present a set of AI techniques designed to look for signatures of new physics in CMB data. In particular, I will introduce how spherical graph neural networks can be applied to perform machine learning directly on CMB temperature and polarization map data. Furthermore, I will discuss how recent advances in agentic AI can assist with a variety of tasks ranging from literature review to scientific code generation and debugging, as well as CMB data analysis. Finally, I will look at a concrete example of employing the aforementioned AI techniques to detect signatures of non-trivial cosmic topology in CMB temperature and polarization data, highlighting my recent work with the COMPACT Collaboration.