Physics-informed CNNs for super-resolution of sparse observations on dynamical systems
File(s)2210.17319v2.pdf (930.61 KB)
Accepted version
Author(s)
Kelshaw, Daniel
Rigas, Georgios
Magri, Luca
Type
Conference Paper
Abstract
In the absence of high-resolution samples, super-resolution of sparse
observations on dynamical systems is a challenging problem with wide-reaching
applications in experimental settings. We showcase the application of
physics-informed convolutional neural networks for super-resolution of sparse
observations on grids. Results are shown for the chaotic-turbulent Kolmogorov
flow, demonstrating the potential of this method for resolving finer scales of
turbulence when compared with classic interpolation methods, and thus
effectively reconstructing missing physics.
observations on dynamical systems is a challenging problem with wide-reaching
applications in experimental settings. We showcase the application of
physics-informed convolutional neural networks for super-resolution of sparse
observations on grids. Results are shown for the chaotic-turbulent Kolmogorov
flow, demonstrating the potential of this method for resolving finer scales of
turbulence when compared with classic interpolation methods, and thus
effectively reconstructing missing physics.
Date Issued
2022-12-03
Date Acceptance
2022-12-01
Citation
2022
Identifier
http://arxiv.org/abs/2210.17319v2
Source
36th conference on Neural Information Processing Systems (NeurIPS)
Subjects
cs.LG
physics.flu-dyn
physics.flu-dyn
Notes
Published in NeurIPS 2022: Machine Learning and the Physical Sciences Workshop. Code at https://github.com/magrilab/pisr. arXiv admin note: text overlap with arXiv:2210.16215
Publication Status
Published
Start Date
2022-12-03
Finish Date
2022-12-03
Coverage Spatial
New Orleans, United States