avatar
AstroAI
Developing Artificial Intelligence to Solve the Mysteries of the Universe
  • HOME
  • RESEARCH
  • EARTHAI
  • PEOPLE
  • EVENTS
  • LATEST NEWS
  • LUNCH TALKS
  • SUMMER PROGRAM
  • WORKSHOP
  • APPLY
  • CONTACT
Home Probing Self-Interacting Dark Matter with Interpretable Neural Networks
Workshop_abstract2024
Cancel

Probing Self-Interacting Dark Matter with Interpretable Neural Networks

Details Invited Speakers Abstracts Register Schedule Venue Accommodations Code of Conduct

Presenter: Ethan Tregidga

Title: Probing Self-Interacting Dark Matter with Interpretable Neural Networks

Date/Time: Tuesday, June 18th, 12:15 PM

Abstract: While the ΛCDM model has proven successful in cosmology, it faces challenges in explaining observations of dwarf galaxies and galaxy clusters. Often, these issues are attributed to unrealistic models of black hole feedback - often known as baryonic feedback. However, it can also be evidence for self-interacting dark matter (SIDM).

Unfortunately, the impact of black holes and SIDM on classical observations of galaxy clusters is degenerate, making it difficult to disentangle one from the other.

Here, we present a semi-supervised clustering neural network that can discern between ΛCDM and SIDM and break the degeneracy between SIDM and baryonic feedback models in galaxy clusters. We leverage data from the BAHAMAS hydro-dynamical simulations, incorporating three models of black hole feedback and four self-interaction cross-sections.

Our neural network achieves comparable classification accuracy to previous methods while yielding an interpretable latent space. This latent space not only incorporates the physicalised cross-section but also captures additional features in the data, such as baryonic feedback effects or simulation nuances, ultimately providing a direct validation check between observations and simulations.

-->

© 2025 AstroAI. Some rights reserved.

Powered by Jekyll with Chirpy theme.

A new version of content is available.