DOE PAGES title logo U.S. Department of Energy
Office of Scientific and Technical Information

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 ( D 3 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 D 3 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 D 3 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:
ORCiD logo [1];  [2];  [3];  [4];  [5];  [6];  [7]
  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:
Research Org.:
Lawrence Berkeley National Laboratory (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

Citation Formats

He, Siyu, Li, Yin, Feng, Yu, Ho, Shirley, Ravanbakhsh, Siamak, Chen, Wei, and Póczos, Barnabás. Learning to predict the cosmological structure formation. United States: N. p., 2019. Web. doi:10.1073/pnas.1821458116.
He, Siyu, Li, Yin, Feng, Yu, Ho, Shirley, Ravanbakhsh, Siamak, Chen, Wei, & Póczos, Barnabás. Learning to predict the cosmological structure formation. United States. https://doi.org/10.1073/pnas.1821458116
He, Siyu, Li, Yin, Feng, Yu, Ho, Shirley, Ravanbakhsh, Siamak, Chen, Wei, and Póczos, Barnabás. Mon . "Learning to predict the cosmological structure formation". United States. https://doi.org/10.1073/pnas.1821458116. https://www.osti.gov/servlets/purl/1561927.
@article{osti_1561927,
title = {Learning to predict the cosmological structure formation},
author = {He, Siyu and Li, Yin and Feng, Yu and Ho, Shirley and Ravanbakhsh, Siamak and Chen, Wei and Póczos, Barnabás},
abstractNote = {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 (D3M), 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 D3M 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 D3M 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.},
doi = {10.1073/pnas.1821458116},
journal = {Proceedings of the National Academy of Sciences of the United States of America},
number = 28,
volume = 116,
place = {United States},
year = {2019},
month = {6}
}

Journal Article:
Free Publicly Available Full Text
Publisher's Version of Record

Citation Metrics:
Cited by: 80 works
Citation information provided by
Web of Science

Figures / Tables:

Fig. 1 Fig. 1: The displacement vector field (Left) and the resulting density field (Right) produced by D3M. The vectors in Left are uniformly scaled down for better visualization.

Save / Share:

Works referenced in this record:

How filaments of galaxies are woven into the cosmic web
journal, April 1996

  • Bond, J. Richard; Kofman, Lev; Pogosyan, Dmitry
  • Nature, Vol. 380, Issue 6575
  • DOI: 10.1038/380603a0

The Baryon Oscillation Spectroscopic Survey of Sdss-Iii
journal, December 2012

  • Dawson, Kyle S.; Schlegel, David J.; Ahn, Christopher P.
  • The Astronomical Journal, Vol. 145, Issue 1
  • DOI: 10.1088/0004-6256/145/1/10

Computing the three-point correlation function of galaxies in $\mathcal {O}(N^2)$ time
journal, October 2015

  • Slepian, Zachary; Eisenstein, Daniel J.
  • Monthly Notices of the Royal Astronomical Society, Vol. 454, Issue 4
  • DOI: 10.1093/mnras/stv2119

Physics 101: Learning Physical Object Properties from Unlabeled Videos
conference, January 2016

  • Wu, Jiajun; Lim, Joseph; Zhang, Hongyi
  • Procedings of the British Machine Vision Conference 2016
  • DOI: 10.5244/C.30.39

A new parallel code for very large-scale cosmological simulations
journal, December 1998


The Sdss-Iv Extended Baryon Oscillation Spectroscopic Survey: Overview and Early data
journal, February 2016

  • Dawson, Kyle S.; Kneib, Jean-Paul; Percival, Will J.
  • The Astronomical Journal, Vol. 151, Issue 2
  • DOI: 10.3847/0004-6256/151/2/44

Deep learning
journal, May 2015

  • LeCun, Yann; Bengio, Yoshua; Hinton, Geoffrey
  • Nature, Vol. 521, Issue 7553
  • DOI: 10.1038/nature14539

Planck 2015 results : X. Diffuse component separation: Foreground maps
journal, September 2016


Mastering the game of Go with deep neural networks and tree search
journal, January 2016

  • Silver, David; Huang, Aja; Maddison, Chris J.
  • Nature, Vol. 529, Issue 7587
  • DOI: 10.1038/nature16961

The Zel'dovich approximation
journal, February 2014

  • White, Martin
  • Monthly Notices of the Royal Astronomical Society, Vol. 439, Issue 4
  • DOI: 10.1093/mnras/stu209

nbodykit: An Open-source, Massively Parallel Toolkit for Large-scale Structure
journal, September 2018


Cosmology and fundamental physics with the Euclid satellite
journal, April 2018

  • Amendola, Luca; Appleby, Stephen; Avgoustidis, Anastasios
  • Living Reviews in Relativity, Vol. 21, Issue 1
  • DOI: 10.1007/s41114-017-0010-3

DeepSphere: Efficient spherical convolutional neural network with HEALPix sampling for cosmological applications
journal, April 2019


Deep Learning
text, January 2018


Newtonian Image Understanding: Unfolding the Dynamics of Objects in Static Images
conference, June 2016

  • Mottaghi, Roozbeh; Bagherinezhad, Hessam; Rastegari, Mohammad
  • 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
  • DOI: 10.1109/CVPR.2016.383

Cosmology and Fundamental Physics with the Euclid Satellite
journal, September 2013

  • Amendola, Luca; Appleby, Stephen; Bacon, David
  • Living Reviews in Relativity, Vol. 16, Issue 1
  • DOI: 10.12942/lrr-2013-6

V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation
conference, October 2016

  • Milletari, Fausto; Navab, Nassir; Ahmadi, Seyed-Ahmad
  • 2016 Fourth International Conference on 3D Vision (3DV)
  • DOI: 10.1109/3DV.2016.79

FastPM: a new scheme for fast simulations of dark matter and haloes
journal, August 2016

  • Feng, Yu; Chu, Man-Yat; Seljak, Uroš
  • Monthly Notices of the Royal Astronomical Society, Vol. 463, Issue 3
  • DOI: 10.1093/mnras/stw2123

CMU DeepLens: deep learning for automatic image-based galaxy–galaxy strong lens finding
journal, July 2017

  • Lanusse, François; Ma, Quanbin; Li, Nan
  • Monthly Notices of the Royal Astronomical Society, Vol. 473, Issue 3
  • DOI: 10.1093/mnras/stx1665

Galaxy And Mass Assembly (GAMA): end of survey report and data release 2
journal, July 2015

  • Liske, J.; Baldry, I. K.; Driver, S. P.
  • Monthly Notices of the Royal Astronomical Society, Vol. 452, Issue 2
  • DOI: 10.1093/mnras/stv1436

Planck 2015 results : XXIII. The thermal Sunyaev-Zeldovich effect-cosmic infrared background correlation
journal, September 2016


Predicting the sequence specificities of DNA- and RNA-binding proteins by deep learning
journal, July 2015

  • Alipanahi, Babak; Delong, Andrew; Weirauch, Matthew T.
  • Nature Biotechnology, Vol. 33, Issue 8
  • DOI: 10.1038/nbt.3300

Bayesian physical reconstruction of initial conditions from large-scale structure surveys
journal, April 2013

  • Jasche, Jens; Wandelt, Benjamin D.
  • Monthly Notices of the Royal Astronomical Society, Vol. 432, Issue 2
  • DOI: 10.1093/mnras/stt449

Planck 2015 results : XIII. Cosmological parameters
journal, September 2016


The 2dF Galaxy Redshift Survey: spectra and redshifts
journal, December 2001


GADGET: a code for collisionless and gasdynamical cosmological simulations
journal, April 2001


Lagrangian theory of gravitational instability of Friedman-Lematre cosmologies - a generic third-order model for non-linear clustering
journal, April 1994


The initial conditions of the Universe from constrained simulations
journal, November 2012

  • Kitaura, Francisco-Shu
  • Monthly Notices of the Royal Astronomical Society: Letters, Vol. 429, Issue 1
  • DOI: 10.1093/mnrasl/sls029

Planck 2015 results : XVI. Isotropy and statistics of the CMB
journal, September 2016


Solving the quantum many-body problem with artificial neural networks
journal, February 2017


Star–galaxy classification using deep convolutional neural networks
journal, October 2016

  • Kim, Edward J.; Brunner, Robert J.
  • Monthly Notices of the Royal Astronomical Society, Vol. 464, Issue 4
  • DOI: 10.1093/mnras/stw2672

U-Net: Convolutional Networks for Biomedical Image Segmentation
book, November 2015

  • Ronneberger, Olaf; Fischer, Philipp; Brox, Thomas
  • Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III
  • DOI: 10.1007/978-3-319-24574-4_28

A volumetric deep Convolutional Neural Network for simulation of mock dark matter halo catalogues
journal, November 2018

  • Berger, Philippe; Stein, George
  • Monthly Notices of the Royal Astronomical Society, Vol. 482, Issue 3
  • DOI: 10.1093/mnras/sty2949

Densely Connected Convolutional Networks
conference, July 2017

  • Huang, Gao; Liu, Zhuang; Maaten, Laurens van der
  • 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
  • DOI: 10.1109/CVPR.2017.243

CosmoFlow: Using Deep Learning to Learn the Universe at Scale
conference, November 2018

  • Mathuriya, Amrita; Bard, Deborah; Mendygral, Peter
  • SC18: International Conference for High Performance Computing, Networking, Storage and Analysis
  • DOI: 10.1109/SC.2018.00068

Long-Term Recurrent Convolutional Networks for Visual Recognition and Description
journal, April 2017

  • Donahue, Jeff; Hendricks, Lisa Anne; Rohrbach, Marcus
  • IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 39, Issue 4
  • DOI: 10.1109/tpami.2016.2599174

Planck 2015 results : XXVI. The Second
journal, September 2016


Photometric Supernova Classification with Machine Learning
journal, August 2016

  • Lochner, Michelle; McEwen, Jason D.; Peiris, Hiranya V.
  • The Astrophysical Journal Supplement Series, Vol. 225, Issue 2
  • DOI: 10.3847/0067-0049/225/2/31

Exploring the posterior surface of the large scale structure reconstruction
journal, July 2018

  • Feng, Yu; Seljak, Uroš; Zaldarriaga, Matias
  • Journal of Cosmology and Astroparticle Physics, Vol. 2018, Issue 07
  • DOI: 10.1088/1475-7516/2018/07/043

Sdss-Iii: Massive Spectroscopic Surveys of the Distant Universe, the Milky way, and Extra-Solar Planetary Systems
journal, August 2011

  • Eisenstein, Daniel J.; Weinberg, David H.; Agol, Eric
  • The Astronomical Journal, Vol. 142, Issue 3
  • DOI: 10.1088/0004-6256/142/3/72

Human-level control through deep reinforcement learning
journal, February 2015

  • Mnih, Volodymyr; Kavukcuoglu, Koray; Silver, David
  • Nature, Vol. 518, Issue 7540
  • DOI: 10.1038/nature14236

Opportunities and obstacles for deep learning in biology and medicine
journal, April 2018

  • Ching, Travers; Himmelstein, Daniel S.; Beaulieu-Jones, Brett K.
  • Journal of The Royal Society Interface, Vol. 15, Issue 141
  • DOI: 10.1098/rsif.2017.0387

Cosmology and fundamental physics with the Euclid satellite
text, January 2018


The 6dF Galaxy Survey: final redshift release (DR3) and southern large-scale structures
journal, October 2009


Fast automated analysis of strong gravitational lenses with convolutional neural networks
journal, August 2017

  • Hezaveh, Yashar D.; Levasseur, Laurence Perreault; Marshall, Philip J.
  • Nature, Vol. 548, Issue 7669
  • DOI: 10.1038/nature23463

Long-term recurrent convolutional networks for visual recognition and description
conference, June 2015

  • Donahue, Jeff; Hendricks, Lisa Anne; Guadarrama, Sergio
  • 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
  • DOI: 10.1109/CVPR.2015.7298878

Simulation as an engine of physical scene understanding
journal, October 2013

  • Battaglia, P. W.; Hamrick, J. B.; Tenenbaum, J. B.
  • Proceedings of the National Academy of Sciences, Vol. 110, Issue 45
  • DOI: 10.1073/pnas.1306572110

Planning chemical syntheses with deep neural networks and symbolic AI
journal, March 2018

  • Segler, Marwin H. S.; Preuss, Mike; Waller, Mark P.
  • Nature, Vol. 555, Issue 7698
  • DOI: 10.1038/nature25978

DeepCMB: Lensing reconstruction of the cosmic microwave background with deep neural networks
journal, July 2019


The evolution of large-scale structure in a universe dominated by cold dark matter
journal, May 1985

  • Davis, M.; Efstathiou, G.; Frenk, C. S.
  • The Astrophysical Journal, Vol. 292
  • DOI: 10.1086/163168

Helmholtz decomposition of the Lagrangian displacement
journal, April 2014


Cosmology and fundamental physics with the Euclid satellite
text, January 2012


The Baryon Oscillation Spectroscopic Survey of SDSS-III
text, January 2012


The Zeldovich approximation
text, January 2014


Long-term Recurrent Convolutional Networks for Visual Recognition and Description
preprint, January 2014


Planck 2015 results. XIII. Cosmological parameters
text, January 2015


The SDSS-IV extended Baryon Oscillation Spectroscopic Survey: Overview and Early Data
text, January 2015


Photometric Supernova Classification With Machine Learning
text, January 2016


V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation
preprint, January 2016


Densely Connected Convolutional Networks
preprint, January 2016


Exploring the posterior surface of the large scale structure reconstruction
text, January 2018


TreePM: A code for cosmological N-body simulations
journal, December 2002

  • Bagla, J. S.
  • Journal of Astrophysics and Astronomy, Vol. 23, Issue 3-4
  • DOI: 10.1007/bf02702282

DeepSphere: Efficient spherical convolutional neural network with HEALPix sampling for cosmological applications
journal, April 2019


AI designs organic syntheses
journal, March 2018


Deep learning
journal, May 2015

  • LeCun, Yann; Bengio, Yoshua; Hinton, Geoffrey
  • Nature, Vol. 521, Issue 7553
  • DOI: 10.1038/nature14539

Fast automated analysis of strong gravitational lenses with convolutional neural networks
journal, August 2017

  • Hezaveh, Yashar D.; Levasseur, Laurence Perreault; Marshall, Philip J.
  • Nature, Vol. 548, Issue 7669
  • DOI: 10.1038/nature23463

Planning chemical syntheses with deep neural networks and symbolic AI
journal, March 2018

  • Segler, Marwin H. S.; Preuss, Mike; Waller, Mark P.
  • Nature, Vol. 555, Issue 7698
  • DOI: 10.1038/nature25978

Predicting the sequence specificities of DNA- and RNA-binding proteins by deep learning
journal, July 2015

  • Alipanahi, Babak; Delong, Andrew; Weirauch, Matthew T.
  • Nature Biotechnology, Vol. 33, Issue 8
  • DOI: 10.1038/nbt.3300

Deep learning for regulatory genomics
journal, August 2015

  • Park, Yongjin; Kellis, Manolis
  • Nature Biotechnology, Vol. 33, Issue 8
  • DOI: 10.1038/nbt.3313

Simulation as an engine of physical scene understanding
journal, October 2013

  • Battaglia, P. W.; Hamrick, J. B.; Tenenbaum, J. B.
  • Proceedings of the National Academy of Sciences, Vol. 110, Issue 45
  • DOI: 10.1073/pnas.1306572110

Exploring the posterior surface of the large scale structure reconstruction
journal, July 2018

  • Feng, Yu; Seljak, Uroš; Zaldarriaga, Matias
  • Journal of Cosmology and Astroparticle Physics, Vol. 2018, Issue 07
  • DOI: 10.1088/1475-7516/2018/07/043

Bayesian physical reconstruction of initial conditions from large-scale structure surveys
journal, April 2013

  • Jasche, Jens; Wandelt, Benjamin D.
  • Monthly Notices of the Royal Astronomical Society, Vol. 432, Issue 2
  • DOI: 10.1093/mnras/stt449

Galaxy And Mass Assembly (GAMA): end of survey report and data release 2
journal, July 2015

  • Liske, J.; Baldry, I. K.; Driver, S. P.
  • Monthly Notices of the Royal Astronomical Society, Vol. 452, Issue 2
  • DOI: 10.1093/mnras/stv1436

Machine learning for quantum physics
journal, February 2017


Ready or Not, Here We Go
journal, June 2016


Cosmology and Fundamental Physics with the Euclid Satellite
journal, September 2013

  • Amendola, Luca; Appleby, Stephen; Bacon, David
  • Living Reviews in Relativity, Vol. 16, Issue 1
  • DOI: 10.12942/lrr-2013-6

Computing the Three-Point Correlation Function of Galaxies in $\mathcal{O}(N^2)$ Time
text, January 2015


Learning Visual Predictive Models of Physics for Playing Billiards
preprint, January 2015


Learning Physical Intuition of Block Towers by Example
preprint, January 2016


Accelerating Eulerian Fluid Simulation With Convolutional Networks
preprint, January 2016


A New Parallel P3M Code for Very Large-Scale Cosmological Simulations
text, January 1998


Works referencing / citing this record:

Toward Improved Predictions in Ungauged Basins: Exploiting the Power of Machine Learning
journal, December 2019

  • Kratzert, Frederik; Klotz, Daniel; Herrnegger, Mathew
  • Water Resources Research, Vol. 55, Issue 12
  • DOI: 10.1029/2019wr026065

A Hybrid Deep Learning Approach to Cosmological Constraints from Galaxy Redshift Surveys
journal, February 2020

  • Ntampaka, Michelle; Eisenstein, Daniel J.; Yuan, Sihan
  • The Astrophysical Journal, Vol. 889, Issue 2
  • DOI: 10.3847/1538-4357/ab5f5e

Painting halos from cosmic density fields of dark matter with physically motivated neural networks
journal, August 2019


Unveiling the predictive power of static structure in glassy systems
journal, April 2020


Painting halos from cosmic density fields of dark matter with physically motivated neural networks
text, January 2019


A Hybrid Deep Learning Approach to Cosmological Constraints From Galaxy Redshift Surveys
text, January 2019


Investigating cosmological GAN emulators using latent space interpolation
journal, July 2021

  • Tamosiunas, Andrius; Winther, Hans A.; Koyama, Kazuya
  • Monthly Notices of the Royal Astronomical Society, Vol. 506, Issue 2
  • DOI: 10.1093/mnras/stab1879

Learning to Simulate Complex Physics with Graph Networks
preprint, January 2020


Fast and Accurate Non-Linear Predictions of Universes with Deep Learning
preprint, January 2020


Figures/Tables have been extracted from DOE-funded journal article accepted manuscripts.