skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.
Attention:The NSF Public Access Repository (NSF-PAR) system and access will be unavailable from 7:00 AM ET to 7:30 AM ET on Friday, April 24 due to maintenance. We apologize for the inconvenience.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: AI-assisted superresolution cosmological simulations
Cosmological simulations of galaxy formation are limited by finite computational resources. We draw from the ongoing rapid advances in artificial intelligence (AI; specifically deep learning) to address this problem. Neural networks have been developed to learn from high-resolution (HR) image data and then make accurate superresolution (SR) versions of different low-resolution (LR) images. We apply such techniques to LR cosmological N-body simulations, generating SR versions. Specifically, we are able to enhance the simulation resolution by generating 512 times more particles and predicting their displacements from the initial positions. Therefore, our results can be viewed as simulation realizations themselves, rather than projections, e.g., to their density fields. Furthermore, the generation process is stochastic, enabling us to sample the small-scale modes conditioning on the large-scale environment. Our model learns from only 16 pairs of small-volume LR-HR simulations and is then able to generate SR simulations that successfully reproduce the HR matter power spectrum to percent level up to 16   h − 1 Mpc and the HR halo mass function to within 10 % down to 1 0 11   M ⊙ . We successfully deploy the model in a box 1,000 times larger than the training simulation box, showing that high-resolution mock surveys can be generated rapidly. We conclude that AI assistance has the potential to revolutionize modeling of small-scale galaxy-formation physics in large cosmological volumes.  more » « less
Award ID(s):
1716131 1909193 2020295 1817256
PAR ID:
10286741
Author(s) / Creator(s):
; ; ; ; ;
Date Published:
Journal Name:
Proceedings of the National Academy of Sciences
Volume:
118
Issue:
19
ISSN:
0027-8424
Page Range / eLocation ID:
e2022038118
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; ; ; ; (, Monthly Notices of the Royal Astronomical Society)
    null (Ed.)
    ABSTRACT In this work, we expand and test the capabilities of our recently developed superresolution (SR) model to generate high-resolution (HR) realizations of the full phase-space matter distribution, including both displacement and velocity, from computationally cheap low-resolution (LR) cosmological N-body simulations. The SR model enhances the simulation resolution by generating 512 times more tracer particles, extending into the deeply nonlinear regime where complex structure formation processes take place. We validate the SR model by deploying the model in 10 test simulations of box size 100 h−1 Mpc, and examine the matter power spectra, bispectra, and two-dimensional power spectra in redshift space. We find the generated SR field matches the true HR result at per cent level down to scales of k ∼ 10 h  Mpc−1. We also identify and inspect dark matter haloes and their substructures. Our SR model generates visually authentic small-scale structures that cannot be resolved by the LR input, and are in good statistical agreement with the real HR results. The SR model performs satisfactorily on the halo occupation distribution, halo correlations in both real and redshift space, and the pairwise velocity distribution, matching the HR results with comparable scatter, thus demonstrating its potential in making mock halo catalogues. The SR technique can be a powerful and promising tool for modelling small-scale galaxy formation physics in large cosmological volumes. 
    more » « less
  2. ; ; ; ; (, Physical Review D)
    High-resolution (HR) simulations in cosmology, in particular when including baryons, can take millions of CPU hours. On the other hand, low-resolution (LR) dark matter simulations of the same cosmological volume use minimal computing resources. We develop a denoising diffusion superresolution emulator for large cosmological simulation volumes. Our approach is based on the image-to-image Palette diffusion model, which we modify to 3 dimensions. Our superresolution emulator is trained to perform outpainting, and can thus upgrade very large cosmological volumes from LR to HR using an iterative outpainting procedure. As an application, we generate a simulation box with 8 times the volume of the Illustris TNG300 training data, constructed with over 9000 outpainting iterations, and quantify its accuracy using various summary statistics. 
    more » « less
  3. ; ; ; ; (, Journal of Geophysical Research: Oceans)
    Abstract Emerging high‐resolution global ocean climate models are expected to improve both hindcasts and forecasts of coastal sea level variability by better resolving ocean turbulence and other small‐scale phenomena. To examine this hypothesis, we compare annual to multidecadal coastal sea level variability over the 1993–2018 period, as observed by tide gauges and as simulated by two identically forced ocean models, at (LR) and (HR) horizontal resolution. Differences between HR and LR, and misfits with tide gauges, are spatially coherent at regional alongcoast scales. Resolution‐related improvements are largest in, and near, marginal seas. Near attached western boundary currents, sea level variance is several times greater in HR than LR, but correlations with observations may be reduced, due to intrinsic ocean variability. Globally, in HR simulations, intrinsic variability comprises from zero to over 80% of coastal sea level variance. Outside of eddy‐rich regions, simulated coastal sea level variability is generally damped relative to observations. We hypothesize that weak coastal variability is related to large‐scale, remotely forced, variability; in both HR and LR, tropical sea level variance is underestimated by 50% relative to satellite altimetric observations. Similar coastal dynamical regimes (e.g., attached western boundary currents) exhibit a consistent sensitivity to horizontal resolution, suggesting that these findings are generalizable to regions with limited coastal observations. 
    more » « less
  4. ; ; ; ; ; ; ; ; ; ; et al (, Monthly Notices of the Royal Astronomical Society)
    ABSTRACT Using high-resolution cosmological radiation-hydrodynamic (RHD) simulations (thesan-hr), we explore the impact of alternative dark matter (altDM) models on galaxies during the Epoch of Reionization. The simulations adopt the IllustrisTNG galaxy formation model. We focus on altDM models that exhibit small-scale suppression of the matter power spectrum, namely warm dark matter (WDM), fuzzy dark matter (FDM), and interacting dark matter (IDM) with strong dark acoustic oscillations (sDAO). In altDM scenarios, both the halo mass functions and the ultraviolet luminosity functions at z ≳ 6 are suppressed at the low-mass/faint end, leading to delayed global star formation and reionization histories. However, strong non-linear effects enable altDM models to ‘catch up’ with cold dark matter (CDM) in terms of star formation and reionization. The specific star formation rates are enhanced in halos below the half-power mass in altDM models. This enhancement coincides with increased gas abundance, reduced gas depletion times, more compact galaxy sizes, and steeper metallicity gradients at the outskirts of the galaxies. These changes in galaxy properties can help disentangle altDM signatures from a range of astrophysical uncertainties. Meanwhile, it is the first time that altDM models have been studied in RHD simulations of galaxy formation. We uncover significant systematic uncertainties in reionization assumptions on the faint-end luminosity function. This underscores the necessity of accurately modeling the small-scale morphology of reionization in making predictions for the low-mass galaxy population. Upcoming James Webb Space Telescope imaging surveys of deep lensed fields hold potential for uncovering the faint low-mass galaxy population, which could provide constraints on altDM models. 
    more » « less
  5. ; ; (, Proceedings of the 29th ACM International Conference on Multimedia (MM'21))
    Super-resolution (SR) is a well-studied technique for reconstructing high-resolution (HR) images from low-resolution (LR) ones. SR holds great promise for video streaming since an LR video segment can be transmitted from the video server to the client that then reconstructs the HR version using SR, resulting in a significant reduction in network bandwidth. However, SR is seldom used in practice for real-time video streaming, because the computational overhead of frame reconstruction results in large latency and low frame rate. To reduce the computational overhead and make SR practical, we propose a deep-learning-based SR method called Fo veated Cas caded Video Super Resolution (focas). focas relies on the fact that human eyes only have high acuity in a tiny central foveal region of the retina. focas uses more neural network blocks in the foveal region to provide higher video quality, while using fewer blocks in the periphery as lower quality is sufficient. To optimize the computational resources and reduce reconstruction latency, focas formulates and solves a convex optimization problem to decide the number of neural network blocks to use in each region of the frame. Using extensive experiments, we show that focas reduces the latency by 50%-70% while maintaining comparable visual quality as traditional (non-foveated) SR. Further, focas provides a 12-16x reduction in the client-to-server network bandwidth in comparison with sending the full HR video segments. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Text Encoded Conversion
  • XML
  • Save / Share this Record
  • Send to Email