An official website of the United States government
In this paper, we investigate and design multiscale simulations for stochastic multiscale PDEs. As for the space, we consider a coarse grid and a known multiscale method, the generalized multiscale finite element method (GMsFEM). In order to obtain a small dimensional representation of the solution in each coarse block, the uncertainty space needs to be partitioned (coarsened). This coarsenining collects realizations that provide similar multiscale features as outlined in GMsFEM (or other method of choice). This step is known to be computationally demanding as it requires many local solves and clustering based on them. In this work, we take a different approach and learn coarsening the uncertainty space. Our methods use deep learning techniques in identifying clusters (coarsening) in the uncertainty space. We use convolutional neural networks combined with some techniques in adversary neural networks. We define appropriate loss functions in the proposed neural networks, where the loss function is composed of several parts that includes terms related to clusters and reconstruction of basis functions. We present numerical results for channelized permeability fields in the examples of flows in porous media.
more »
« less
MINER: Multiscale Implicit Neural Representation
We introduce a new neural signal model designed for efficient high-resolution representation of large-scale signals. The key innovation in our multiscale implicit neural representation (MINER) is an internal representation via a Laplacian pyramid, which provides a sparse multiscale decomposition of the signal that captures orthogonal parts of the signal across scales. We leverage the advantages of the Laplacian pyramid by representing small disjoint patches of the pyramid at each scale with a small MLP. This enables the capacity of the network to adaptively increase from coarse to fine scales, and only represent parts of the signal with strong signal energy. The parameters of each MLP are optimized from coarse-to-fine scale which results in faster approximations at coarser scales, thereby ultimately an extremely fast training process. We apply MINER to a range of large-scale signal representation tasks, including gigapixel images and very large point clouds, and demonstrate that it requires fewer than 25% of the parameters, 33% of the memory footprint, and 10% of the computation time of competing techniques such as ACORN to reach the same representation accuracy.
more »
« less
- PAR ID:
- 10466354
- Date Published:
- Journal Name:
- European Conference on Computer Vision (ECCV) 2022
- Page Range / eLocation ID:
- 318–333
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Surface Variability Mapping and Roughness Analysis of the Moon Using a Coarse‐Graining DecompositionAbstract The lunar surface contains a wide variety of topographic shapes and features, each with different distributions and scales, and any analysis technique to objectively measure roughness must respect these qualities. Coarse‐graining is a naturally scale‐dependent filtering technique that preserves scale‐dependent symmetries and produces coarse elevation maps that gradually erase the smaller features from the original topography. In this study of the lunar surface, we present two surface variability metrics obtained from coarse‐graining lunar topography: fine elevation and coarse curvature. Both metrics are isotropic, deterministic, slope‐independent, and coordinate‐agnostic. Fine (detrended) elevation is acquired by subtracting the coarse elevation from the original topography and contains features that are smaller than the coarse‐graining length‐scale. Coarse curvature is the Laplacian of coarsened topography, and naturally quantifies the curvature at any scale and indicates whether a location is elevated or depressed relative to its neighborhood at that scale. We find that highlands and maria have distinct roughness characteristics at all length‐scales. Our topographic spectra reveal four scale‐breaks that mark characteristic shifts in surface roughness: 100, 300, 1,000, and 4,000 km. Comparing fine elevation distributions between maria and highlands, we show that maria fine elevation is biased toward smaller‐magnitude elevations and that the maria–highland discrepancies are more pronounced at larger length‐scales. We also provide local examples of selected regions to demonstrate that these metrics can successfully distinguish geological features of different length‐scales.more » « less
-
This paper presents a multiscale modeling framework (MMF) to model moist at- mospheric limited-area weather. The MMF resolves large-scale convection using a coarse grid while simultaneously resolving local features through numerous fine local grids and coupling them seamlessly. Both large- and small-scale processes are modeled using the compressible Navier-Stokes equations within the Nonhydrostatic Unified Model of the Atmosphere (NUMA), and they are discretized using a continuous element-based Galerkin method (spectral elements) with high-order basis functions. Consequently, the large-scale and small-scale models share the same dynamical core but have the flexibility to be ad- justed individually. The proposed MMF method is tested in 2D and 3D idealized limited- area weather problems involving storm clouds produced by squall line and supercell sim- ulations. The MMF numerical results showed enhanced representation of cloud processes compared to the coarse model.more » « less
-
The functional connectome supports information transmission through the brain at various spatial scales, from exchange between broad cortical regions to finer-scale, vertex-wise connections that underlie specific information processing mechanisms. In adults, while both the coarse- and fine-scale functional connectomes predict cognition, the fine scale can predict up to twice the variance as the coarse-scale functional connectome. Yet, past brain-wide association studies, particularly using large developmental samples, focus on the coarse connectome to understand the neural underpinnings of individual differences in cognition. Using a large cohort of children (age 9–10 years;n = 1,115 individuals; both sexes; 50% female, including 170 monozygotic and 219 dizygotic twin pairs and 337 unrelated individuals), we examine the reliability, heritability, and behavioral relevance of resting-state functional connectivity computed at different spatial scales. We use connectivity hyperalignment to improve access to reliable fine-scale (vertex-wise) connectivity information and compare the fine-scale connectome with the traditional parcel-wise (coarse scale) functional connectomes. Though individual differences in the fine-scale connectome are more reliable than those in the coarse-scale, they are less heritable. Further, the alignment and scale of connectomes influence their ability to predict behavior, whereby some cognitive traits are equally well predicted by both connectome scales, but other, less heritable cognitive traits are better predicted by the fine-scale connectome. Together, our findings suggest there are dissociable individual differences in information processing represented at different scales of the functional connectome which, in turn, have distinct implications for heritability and cognition.more » « less
-
null (Ed.)A turbulence enrichment model for subfilter-scale motions in large eddy simulations (LES) is comprehensively evaluated in the context of a posteriori analysis. The paper further develops the Gabor mode enrichment model first introduced in Ghate & Lele (J. Fluid Mech., vol. 819, 2017, pp. 494–539) by analysing three key requisites of LES enrichment using solenoidal small-scale velocity fields: (a) consistent spectral extrapolation and improvement of resolved single- and two-point second-order correlations; (b) ability to accurately capture the flow physics responsible for temporal decorrelation at small scales; and (c) accurate representation of spatially localized and intermittent interscale energy transfer between scales resolved by the coarse-grid LES and subfilter scales. We argue that the spatially and spectrally localized Gabor wavepackets offer an optimal basis to represent small-scale turbulence within quasi-homogeneous regions, although the alignment of fine-scale vorticity with large-scale strain appears to be somewhat overemphasized. Consequently, we interpret the resulting subfilter scales as those induced by a set of spatially dispersed Burgers–Townsend vortices with orientations determined by the larger scale velocity gradients resolved by the coarse-grid LES. Enrichment of coarse-grid simulations of two high Reynolds number flow configurations, homogeneous isotropic turbulence and a rough-wall turbulent boundary layer show promising results.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
