Dark Matter Subhalos, Strong Lensing and Machine Learning

TL;DR

This paper explores using machine learning, specifically convolutional neural networks, to analyze strong lensing images for detecting low-mass dark matter sub-halos, achieving high accuracy in classification.

Contribution

It introduces a novel application of CNNs to classify dark matter sub-halo mass cut-offs in lensing images with high precision.

Findings

CNNs correctly identify mass cut-offs within an order of magnitude

Achieved over 93% classification accuracy

Demonstrates potential for dark matter substructure detection

Abstract

We investigate the possibility of applying machine learning techniques to images of strongly lensed galaxies to detect a low mass cut-off in the spectrum of dark matter sub-halos within the lens system. We generate lensed images of systems containing substructure in seven different categories corresponding to lower mass cut-offs ranging from $10^9M_\odot$ down to $10^6M_\odot$. We use convolutional neural networks to perform a multi-classification sorting of these images and see that the algorithm is able to correctly identify the lower mass cut-off within an order of magnitude to better than 93% accuracy.