Title: Learning to predict the cosmological structure formation

Abstract

Matter evolved under the influence of gravity from minuscule density fluctuations. Nonperturbative structure formed hierarchically over all scales and developed non-Gaussian features in the Universe, known as the cosmic web. To fully understand the structure formation of the Universe is one of the holy grails of modern astrophysics. Astrophysicists survey large volumes of the Universe and use a large ensemble of computer simulations to compare with the observed data to extract the full information of our own Universe. However, to evolve billions of particles over billions of years, even with the simplest physics, is a daunting task. We build a deep neural network, the Deep Density Displacement Model ( ${\mathrm{D}}^3\mathrm{M}$), which learns from a set of prerun numerical simulations, to predict the nonlinear large-scale structure of the Universe with the Zel’dovich Approximation (ZA), an analytical approximation based on perturbation theory, as the input. Our extensive analysis demonstrates that ${\mathrm{D}}^3\mathrm{M}$ outperforms the second-order perturbation theory (2LPT), the commonly used fast-approximate simulation method, in predicting cosmic structure in the nonlinear regime. We also show that ${\mathrm{D}}^3\mathrm{M}$ is able to accurately extrapolate far beyond its training data and predict structure formation for significantly different cosmological parameters. Our study proves that deep learning is a practical and accurate alternative to approximate 3D simulations of the gravitational structure formation of the Universe.

Authors:

He, Siyu  ^[1];

• Search DOE PAGES for author "He, Siyu"
• Search DOE PAGES for ORCID "0000-0002-1058-4105"
• Search orcid.org for ORCID "0000-0002-1058-4105"

Li, Yin ^[2]; Feng, Yu ^[3]; Ho, Shirley ^[4]; Ravanbakhsh, Siamak ^[5]; Chen, Wei ^[6]; Póczos, Barnabás ^[7]

---

+ Show Author Affiliations

1. Carnegie Mellon Univ., Pittsburgh, PA (United States); Flatiron Inst. New York, NY (United States); Univ. of Tokyo (Japan)
2. Univ. of California, Berkeley, CA (United States); Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States); Univ. of Tokyo (Japan)
3. Univ. of California, Berkeley, CA (United States); Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States)
4. Carnegie Mellon Univ., Pittsburgh, PA (United States); Flatiron Inst. New York, NY (United States); Univ. of Tokyo (Japan); Univ. of California, Berkeley, CA (United States); Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States)
5. Univ. of British Columbia, Vancouver, BC (Canada)
6. Flatiron Inst. New York, NY (United States)
7. Carnegie Mellon Univ., Pittsburgh, PA (United States)

Publication Date:

2019-06-24

Research Org.:

Lawrence Berkeley National Lab. (LBNL), Berkeley, CA (United States)

Sponsoring Org.:

USDOE Office of Science (SC); National Aeronautics and Space Administration (NASA); Simons Foundation

OSTI Identifier:

1561927

Grant/Contract Number:  

AC02-05CH11231; 15-WFIRST15-0008; 12-EUCLID12-0004

Resource Type:

Accepted Manuscript

Journal Name:

Proceedings of the National Academy of Sciences of the United States of America

Additional Journal Information:

Journal Volume: 116; Journal Issue: 28; Journal ID: ISSN 0027-8424

Publisher:

National Academy of Sciences, Washington, DC (United States)

Country of Publication:

United States

Language:

English

Subject:

79 ASTRONOMY AND ASTROPHYSICS; cosmology; deep learning; simulation