DeepVoid: A Deep Learning Void Detector

Abstract

We present DeepVoid, an application of deep learning trained on a physical definition of cosmic voids to detect voids in density fields and galaxy distributions. By semantically segmenting the IllustrisTNG simulation volume using the tidal tensor, we train a deep convolutional neural network to classify local structure using a U-Net architecture for training and prediction. The model achieves a void F1 score of 0.96 and a Matthews correlation coefficient over all structural classes of 0.81 for dark matter particles in IllustrisTNG with interparticle spacing of $\lambda =0.33h^{-1}\text{Mpc}$. We then apply the machine learning technique of curricular learning to enable the model to classify structure in data with significantly larger intertracer separation. At the highest tracer separation tested, $\lambda =10h^{-1}\text{Mpc}$, the model achieves a void F1 score of 0.89 and a Matthews correlation coefficient of 0.6 on IllustrisTNG subhalos.

23 pages, 9 figures. Submitted to the Astrophysical Journal on 4/25/25. Accepted on 12/11/2025. Published on 02/04/2026

References in corpus (70)

• Planck 2015 results. XIII. Cosmological parameters
• The Seventh Data Release of the Sloan Digital Sky Survey
• Populating a cluster of galaxies - I. Results at z=0
• E pur si muove: Galiliean-invariant cosmological hydrodynamical simulations on a moving mesh
• Introducing the Illustris Project: Simulating the coevolution of dark and visible matter in the Universe
• The Baryon Oscillation Spectroscopic Survey of SDSS-III
• First results from the IllustrisTNG simulations: matter and galaxy clustering
• First results from the IllustrisTNG simulations: the stellar mass content of groups and clusters of galaxies
• First results from the IllustrisTNG simulations: the galaxy color bimodality
• First results from the IllustrisTNG simulations: A tale of two elements -- chemical evolution of magnesium and europium
• First results from the IllustrisTNG simulations: radio haloes and magnetic fields
• Euclid. I. Overview of the Euclid mission
• Properties of Dark Matter Haloes in Clusters, Filaments, Sheets and Voids
• A hierarchy of voids: Much ado about nothing
• ZOBOV: a parameter-free void-finding algorithm
• Tracing the cosmic web
• A Dynamical Classification of the Cosmic Web
• An Imprint of Super-Structures on the Microwave Background due to the Integrated Sachs-Wolfe Effect
• A Cosmic Watershed: the WVF Void Detection Technique
• DESI Bright Galaxy Survey: Final Target Selection, Design, and Validation
• The Multiscale Morphology Filter: Identifying and Extracting Spatial Patterns in the Galaxy Distribution
• Cosmic Voids in Sloan Digital Sky Survey Data Release 7
• Target Selection and Validation of DESI Luminous Red Galaxies
• Voids in the PSCz Survey and the Updated Zwicky Catalog
• Precision cosmography with stacked voids
• Learning to Predict the Cosmological Structure Formation
• The Aspen--Amsterdam Void Finder Comparison Project
• Target Selection and Validation of DESI Quasars
• A kinematic classification of the cosmic web
• Target Selection and Validation of DESI Emission Line Galaxies
• Counting voids to probe dark energy
• Voids in massive neutrino cosmologies
• The Hierarchical Structure and Dynamics of Voids
• Beyond BAO: improving cosmological constraints from BOSS with measurement of the void-galaxy cross-correlation
• The High Latitude Spectroscopic Survey on the Nancy Grace Roman Space Telescope
• Simulating nonlinear cosmological structure formation with massive neutrinos
• ORIGAMI: Delineating Halos using Phase-Space Folds
• Reconstructing the Cosmic Velocity and Tidal Fields with Galaxy Groups Selected from the Sloan Digital Sky Survey
• Fast cosmic web simulations with generative adversarial networks
• A robust public catalogue of voids and superclusters in the SDSS Data Release 7 galaxy surveys
• The Void Size Function in Dynamical Dark Energy Cosmologies
• Precision cosmology with voids in the final BOSS data
• Galaxy And Mass Assembly (GAMA): The galaxy luminosity function within the cosmic web
• Simulating Voids
• Cosmological Reconstruction From Galaxy Light: Neural Network Based Light-Matter Connection
• Alignments of Voids in the Cosmic Web
• On the linearity of tracer bias around voids
• The lensing and temperature imprints of voids on the Cosmic Microwave Background
• Void Lensing as a Test of Gravity
• Testing cosmology with a catalogue of voids in the BOSS galaxy surveys
• Euclid: Cosmological forecasts from the void size function
• Classifying the Large Scale Structure of the Universe with Deep Neural Networks
• Cosmological exploitation of the size function of cosmic voids identified in the distribution of biased tracers
• Cosmic voids in modified gravity scenarios
• The Integrated Sachs-Wolfe effect in $f(R)$ gravity
• Cosmological constraints from the BOSS DR12 void size function
• The nature of voids: II. Tracing underdensities with biased galaxies
• A volumetric deep Convolutional Neural Network for simulation of mock dark matter halo catalogues
• Field Level Neural Network Emulator for Cosmological N-body Simulations
• An interpretable machine learning framework for dark matter halo formation
• The part and the whole: voids, supervoids, and their ISW imprint
• Halo abundances within the cosmic web
• DIVE in the cosmic web: voids with Delaunay Triangulation from discrete matter tracer distributions
• Cosmology with cosmic web environments I. Real-space power spectra
• Updated void catalogs of the SDSS DR7 main sample
• Observational Constraints on Dynamical Dark Energy Models
• Guess the cheese flavour by the size of its holes: A cosmological test using the abundance of Popcorn voids
• Velocity profiles of matter and biased tracers around voids
• Classification of cosmic structures for galaxies with deep learning: connecting cosmological simulations with observations
• A Comparison of Void-Finding Algorithms using Crossing Numbers