Astroparticle Physics Seminar

Insights into dark matter halo density profiles with neural networks

by Dr Luisa Lucie-Smith (MPA)

Europe/Berlin
https://mppmu.zoom.us/j/99300254726 (Main)

https://mppmu.zoom.us/j/99300254726

Main

Via Zoom: https://mppmu.zoom.us/j/99300254726
Description

The density structure of dark matter halos contains key information about cosmology and the nature of dark matter. Numerical simulations have shown that the density profiles of halos exhibit a self-similar functional form for a large range of halo masses and for several different cosmological models. However, the physical origin of this near-universal shape is still not well understood. The lack of a consensus on the origin of self-similar density profiles means that the modeling of profiles relies primarily on empirically-found fitting formulae. I will present two different machine-learning frameworks that aim to provide insights into dark matter halo density profiles. The first focuses on the origin of dark matter density profiles. The goal is to measure the amount of information retained by the final profiles about the haloes' initial conditions and their mass accretion history. The second framework aims to discover what the independent degrees of freedom are in dark matter halo density profiles. I will show that interpretable machine learning frameworks, combined with the information-theoretical metrics such as mutual information, allow us to extract knowledge about the underlying physics from the neural network. 

Organised by

Sajad Abbar