Computer Methods in Applied Mechanics and Engineering, volume 372, pages 113291

Attention-based Convolutional Autoencoders for 3D-Variational Data Assimilation

Julian Mack ¹

Rossella Arcucci ¹

Miguel Molina-Solana ^{2, 3}

Yike Guo ¹

Hide authors affiliations Show authors affiliations: 3 affiliations

Data Science Institute, Imperial College, London, UK |

Data Science Institute, Imperial College London, UK |

Dept. Computer Science and AI, Universidad de Granada, Spain |

Publication type: Journal Article

Publication date: 2020-12-01

Elsevier

Journal: Computer Methods in Applied Mechanics and Engineering

SJR: 2.397

CiteScore: 12.7

Impact factor: 6.9

ISSN: 03742830, 00457825

DOI: 10.1016/j.cma.2020.113291

Copy DOI

Computer Science Applications

General Physics and Astronomy

Mechanical Engineering

Mechanics of Materials

Computational Mechanics

Abstract

We propose a new 'Bi-Reduced Space' approach to solving 3D Variational Data Assimilation using Convolutional Autoencoders. We prove that our approach has the same solution as previous methods but has significantly lower computational complexity; in other words, we reduce the computational cost without affecting the data assimilation accuracy. We tested the new method with data from a real-world application: a pollution model of a site in Elephant and Castle, London and found that we could reduce the size of the background covariance matrix representation by O(10^3) and, at the same time, increase our data assimilation accuracy with respect to existing reduced space methods.

Found

Top-30

Journals

Publishers

	1 2 3 4 5 6 7 8 9
Elsevier	Elsevier, 9, 40.91% Elsevier 9 publications, 40.91%
Wiley	Wiley, 3, 13.64% Wiley 3 publications, 13.64%
AIP Publishing	AIP Publishing, 2, 9.09% AIP Publishing 2 publications, 9.09%
Copernicus	Copernicus, 1, 4.55% Copernicus 1 publication, 4.55%
IOP Publishing	IOP Publishing, 1, 4.55% IOP Publishing 1 publication, 4.55%
Taylor & Francis	Taylor & Francis, 1, 4.55% Taylor & Francis 1 publication, 4.55%
Institute of Electrical and Electronics Engineers (IEEE)	Institute of Electrical and Electronics Engineers (IEEE), 1, 4.55% Institute of Electrical and Electronics Engineers (IEEE) 1 publication, 4.55%
Springer Nature	Springer Nature, 1, 4.55% Springer Nature 1 publication, 4.55%
	1 2 3 4 5 6 7 8 9

We do not take into account publications without a DOI.
Statistics recalculated only for publications connected to researchers, organizations and labs registered on the platform.
Statistics recalculated weekly.

Are you a researcher?

Create a profile to get free access to personal recommendations for colleagues and new articles.

Metrics

Cite this

GOST |

Cite this

GOST Copy

Mack J. et al. Attention-based Convolutional Autoencoders for 3D-Variational Data Assimilation // Computer Methods in Applied Mechanics and Engineering. 2020. Vol. 372. p. 113291.

GOST all authors (up to 50) Copy

Mack J., Arcucci R., Molina-Solana M., Guo Y. Attention-based Convolutional Autoencoders for 3D-Variational Data Assimilation // Computer Methods in Applied Mechanics and Engineering. 2020. Vol. 372. p. 113291.

RIS |

Cite this

RIS Copy

TY - JOUR

DO - 10.1016/j.cma.2020.113291

UR - https://doi.org/10.1016/j.cma.2020.113291

TI - Attention-based Convolutional Autoencoders for 3D-Variational Data Assimilation

T2 - Computer Methods in Applied Mechanics and Engineering

AU - Mack, Julian

AU - Arcucci, Rossella

AU - Molina-Solana, Miguel

AU - Guo, Yike

PY - 2020

DA - 2020/12/01

PB - Elsevier

SP - 113291

VL - 372

SN - 0374-2830

SN - 0045-7825

ER -

BibTex

Cite this

BibTex (up to 50 authors) Copy

@article{2020_Mack,

author = {Julian Mack and Rossella Arcucci and Miguel Molina-Solana and Yike Guo},

title = {Attention-based Convolutional Autoencoders for 3D-Variational Data Assimilation},

journal = {Computer Methods in Applied Mechanics and Engineering},

year = {2020},

volume = {372},

publisher = {Elsevier},

month = {dec},

url = {https://doi.org/10.1016/j.cma.2020.113291},

pages = {113291},

doi = {10.1016/j.cma.2020.113291}

}

Found error?

Publisher

Elsevier

Journal

Computer Methods in Applied Mechanics and Engineering

SJR

2.397

CiteScore

12.7

Impact factor

6.9

ISSN

03742830 (Print)

00457825 (Print)

	1 2 3
Quarterly Journal of the Royal Meteorological Society	Quarterly Journal of the Royal Meteorological Society, 3, 13.64% Quarterly Journal of the Royal Meteorological Society 3 publications, 13.64%
Physics of Fluids	Physics of Fluids, 2, 9.09% Physics of Fluids 2 publications, 9.09%
Journal of Computational Science	Journal of Computational Science, 2, 9.09% Journal of Computational Science 2 publications, 9.09%
Geoscientific Model Development	Geoscientific Model Development, 1, 4.55% Geoscientific Model Development 1 publication, 4.55%
Computers and Fluids	Computers and Fluids, 1, 4.55% Computers and Fluids 1 publication, 4.55%
Physica A: Statistical Mechanics and its Applications	Physica A: Statistical Mechanics and its Applications, 1, 4.55% Physica A: Statistical Mechanics and its Applications 1 publication, 4.55%
Environmental Research Letters	Environmental Research Letters, 1, 4.55% Environmental Research Letters 1 publication, 4.55%
Computers in Biology and Medicine	Computers in Biology and Medicine, 1, 4.55% Computers in Biology and Medicine 1 publication, 4.55%
Nuclear Science and Engineering	Nuclear Science and Engineering, 1, 4.55% Nuclear Science and Engineering 1 publication, 4.55%
IEEE/CAA Journal of Automatica Sinica	IEEE/CAA Journal of Automatica Sinica, 1, 4.55% IEEE/CAA Journal of Automatica Sinica 1 publication, 4.55%
Lecture Notes in Computer Science	Lecture Notes in Computer Science, 1, 4.55% Lecture Notes in Computer Science 1 publication, 4.55%
Neural Networks	Neural Networks, 1, 4.55% Neural Networks 1 publication, 4.55%
Theoretical and Applied Mechanics Letters	Theoretical and Applied Mechanics Letters, 1, 4.55% Theoretical and Applied Mechanics Letters 1 publication, 4.55%
Expert Systems with Applications	Expert Systems with Applications, 1, 4.55% Expert Systems with Applications 1 publication, 4.55%
Computer Physics Communications	Computer Physics Communications, 1, 4.55% Computer Physics Communications 1 publication, 4.55%
	1 2 3