Nonlinear 3D cosmic web simulation with heavy-tailed generative adversarial networks

Abstract

Fast and accurate simulations of the nonlinear evolution of the cosmic density field are a major component of many cosmological analyses, but the computational time and storage required to run them can be exceedingly large. For this reason, we use generative adversarial networks (GANs) to learn a compressed representation of the 3D matter density field that is fast and easy to sample, and for the first time show that GANs are capable of generating samples at the level of accuracy of other conventional methods. Using subvolumes from a suite of GADGET-2 N-body simulations, we demonstrate that a deepconvolutional GAN can generate samples that capture both large-and small-scale features of the matter density field, as validated through a variety of n-point statistics. The use of a data scaling that preserves high-density features and a heavy-tailed latent space prior allow us to obtain state of the art results for fast 3D cosmic web generation. In particular, the mean power spectra from generated samples agree to within 5% up to k ¼ 3 and within 10% for k ≤ 5 when compared with N-body simulations, and similar accuracy is obtained for a variety of bispectra. By modeling the latent space with a heavy-tailed prior rather than a standard Gaussian, we better capture sample variance in the high-density voxel PDF and reduce errors in power spectrum and bispectrum covariance on all scales. Furthermore, we show that a conditional GAN can smoothly interpolate between samples conditioned on redshift. Deep generative models, such as the ones described in this work, provide great promise as fast, low-memory, high-fidelity forward models of large-scale structure.

Large-scale dark matter simulations

“…The main idea of GANs is to have two competing neural networks; one of which creates fake data from a random process and another that seeks to distinguish them from those in the training sample. This architecture has shown to produce 2D and 3D fields quantitatively very similar to those obtained by direct numerical simulation (Rodrı ´guez et al 2018;Perraudin et al 2019;Tamosiunas et al 2021;Ullmo et al 2020;Feder et al 2020), even in previously unseen cosmologies (see Fig. 28 and Perraudin et al 2021).…”

Large-scale dark matter simulations

“…The main idea of GANs is to have two competing neural networks; one of which creates fake data from a random process and another that seeks to distinguish them from those in the training sample. This architecture has shown to produce 2D and 3D fields quantitatively very similar to those obtained by direct numerical simulation (Rodríguez et al 2018;Perraudin et al 2019;Tamosiunas et al 2020;Ullmo et al 2020;Feder et al 2020), even in previously unseen cosmologies (Perraudin et al 2020). The architectures used to create fake data can also be combined with existing data from, for instance, low resolution simulations to artificially increase their resolution.…”

High Energy Physics Calorimeter Detector Simulation Using Generative Adversarial Networks With Domain Related Constraints

“…Similarly, scientific simulation is another domain where GAN has shown immense success. Astro-physics [25], [26], micorbiology [27], and material composition [28], etc., are some of the avenues explored.…”