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

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.

Published in Computer Methods in Applied Mechanics and Engineering in Dec 2020

References in corpus (16)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Auto-Encoding Variational Bayes
• Neural Machine Translation by Jointly Learning to Align and Translate
• Sequence to Sequence Learning with Neural Networks
• Striving for Simplicity: The All Convolutional Net
• Variational image compression with a scale hyperprior
• Tutorial on Variational Autoencoders
• Pruning Filters for Efficient ConvNets
• Deep Autoencoding Models for Unsupervised Anomaly Segmentation in Brain MR Images
• Density Modeling of Images using a Generalized Normalization Transformation
• Deep Rewiring: Training very sparse deep networks
• Residual Dense Network for Image Restoration
• A variational approach to Data Assimilation in the Solar Wind
• Deep Learning and Data Assimilation for Real-Time Production Prediction in Natural Gas Wells
• Multi-scale and Context-adaptive Entropy Model for Image Compression
• Extreme Image Coding via Multiscale Autoencoders With Generative Adversarial Optimization

Cited by in corpus (10)

• Bridging observation, theory and numerical simulation of the ocean using Machine Learning
• Latent Space Data Assimilation by using Deep Learning
• An AI-based Domain-Decomposition Non-Intrusive Reduced-Order Model for Extended Domains applied to Multiphase Flow in Pipes
• Applying Convolutional Neural Networks to Data on Unstructured Meshes with Space-Filling Curves
• 3D-Var Data Assimilation using a Variational Autoencoder
• Data Assimilation in the Latent Space of a Neural Network
• Frame invariant neural network closures for Kraichnan turbulence
• Supervised machine learning to estimate instabilities in chaotic systems: estimation of local Lyapunov exponents
• Joint calibration and mapping of satellite altimetry data using trainable variational models
• Variational Data Assimilation with a Learned Inverse Observation Operator