Variational autoencoder for generating realistic $N$-body simulations for dark matter halos

Abstract

In this paper, we present a deep-learning approach to generate synthetic cosmological images by training a convolutional variational autoencoder on two-dimensional dark matter density slices projected from $\Lambda$CDM $N$-body simulations. The model learns a compact latent representation that enables accurate reconstructions and fast generation of new synthetic realizations through a single forward pass through the decoder. We validate the generated fields using cosmology-based summary statistics, focusing on the matter power spectrum and related Fourier space diagnostics, and found good agreement with the reference simulation across the range of scales where the maps exhibit good resolution. Thanks to its low inference cost and stable training target, this variational-autoencoder approach provides a lightweight and reproducible basis for generative modeling of large-scale projected structures and can support downstream tasks such as fast simulation generation and data augmentation.

18 pages, 9 figures

References in corpus (34)

• The cosmological simulation code GADGET-2
• Simulating the joint evolution of quasars, galaxies and their large-scale distribution
• An Introduction to Variational Autoencoders
• Stable clustering, the halo model and nonlinear cosmological power spectra
• Properties of galaxies reproduced by a hydrodynamic simulation
• Dark Energy Survey Year 1 Results: Cosmological Constraints from Galaxy Clustering and Weak Lensing
• Revising the Halofit Model for the Nonlinear Matter Power Spectrum
• Physics-Constrained Deep Learning for High-dimensional Surrogate Modeling and Uncertainty Quantification without Labeled Data
• Enzo: An Adaptive Mesh Refinement Code for Astrophysics
• Principal Component Analysis: A Natural Approach to Data Exploration
• Reconstruction of dark energy and expansion dynamics using Gaussian processes
• Simulating cosmic structure formation with the GADGET-4 code
• Solving Large Scale Structure in Ten Easy Steps with COLA
• The CAMELS project: Cosmology and Astrophysics with MachinE Learning Simulations
• Bayesian parameter estimation using conditional variational autoencoders for gravitational-wave astronomy
• Mock galaxy catalogs using the quick particle mesh method
• AI-assisted super-resolution cosmological simulations
• Gaussian processes reconstruction of dark energy from observational data
• FlowPM: Distributed TensorFlow Implementation of the FastPM Cosmological N-body Solver
• A deep learning model to emulate simulations of cosmic reionization
• Predicting dark matter halo formation in N-body simulations with deep regression networks
• Neural network reconstructions for the Hubble parameter, growth rate and distance modulus
• Power spectrum for the small-scale Universe
• Hunting down systematics in baryon acoustic oscillations after cosmic high noon
• Encoding large scale cosmological structure with Generative Adversarial Networks
• LINNA: Likelihood Inference Neural Network Accelerator
• Neural Networks Optimized by Genetic Algorithms in Cosmology
• Non-parametric reconstruction of cosmological observables using Gaussian Processes Regression
• Machine learning unveils the linear matter power spectrum of modified gravity
• Deep Learning and genetic algorithms for cosmological Bayesian inference speed-up
• Analysis of Dark Matter Halo Structure Formation in $N$-body Simulations with Machine Learning
• Dark Matter with N-Body Numerical Simulations
• Gravitational Turbulence: the Small-Scale Limit of the Cold-Dark-Matter Power Spectrum
• Accelerating lensed quasar discovery and modeling with physics-informed variational autoencoders