Deep learning the astrometric signature of dark matter substructure

2020/08/31 by Kyriakos Vattis, Michael W. Toomey, Savvas M. Koushiappas
Mathematics · Physics and Astronomy · #Artificial intelligence #Astronomy #Astrophysics #Computer science #Convolutional neural network #Dark Matter and Cosmic Phenomena #Dark matter #Galaxies: Formation, Evolution, Phenomena #Galaxy #Geometry #Gravitational microlensing #Mathematics #Milky Way #Physics #Population #Quasar #Radio Astronomy Observations and Technology #Signature (topology) #Substructure #astro-ph.CO #astro-ph.GA

paper · pdf · doi:10.1103/physrevd.104.123541

Replaced with version accepted for publication in Physics. Rev. D

arxiv created 2021/11/23 · openalex publication_date 2021/12/20 · arxiv updated 2022/01/05 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/06

Abstract

We study the application of machine learning techniques for the detection of the astrometric signature of dark matter substructure. In this proof of principle, a population of dark matter subhalos in the Milky Way will act as lenses for sources of extragalactic origin such as quasars. We train resnet-18, a state-of-the-art convolutional neural network to classify angular velocity maps of a population of quasars into lensed and nonlensed classes. We show that an SKA-like survey with extended operational baseline can be used to probe the substructure content of the Milky Way and demonstrate how axiomatic attribution can be used to localize substructures in lensing maps.

Citations

• Comprehensive search for dark matter annihilation in dwarf galaxies
• The APOSTLE simulations: solutions to the Local Group's cosmic puzzles
• Experimental Searches for the Axion and Axion-Like Particles
• Stream-subhalo interactions in the Aquarius simulations
• Results from a Search for Dark Matter in the Complete LUX Exposure
• Axiomatic Attribution for Deep Networks
• Uncertainties in Parameters Estimated with Neural Networks: Application to Strong Gravitational Lensing
• Machine learning and the physical sciences
• Search for neutrinos from annihilation of captured low-mass dark matter particles in the Sun by Super-Kamiokande
• Deep Residual Learning for Image Recognition
• Power of halometry
• Dark matter and the XENON1T electron recoil excess
• Galactic origin of relativistic bosons and XENON1T excess
• Baryon density extraction and isotropy analysis of cosmic microwave background using deep learning
• Planck 2018 results: VI. Cosmological parameters