DeepAI AI Chat
Log In Sign Up

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

by   Julian Mack, et al.

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.


page 15

page 16

page 29

page 30

page 31

page 33


Physics Guided Machine Learning for Variational Multiscale Reduced Order Modeling

We propose a new physics guided machine learning (PGML) paradigm that le...

Sparsity in Variational Autoencoders

Working in high-dimensional latent spaces, the internal encoding of data...

Cost-effective Variational Active Entity Resolution

Accurately identifying different representations of the same real-world ...

Preconditioning Kaczmarz method by sketching

We propose a new method for preconditioning Kaczmarz method by sketching...

Deep Clustering of Compressed Variational Embeddings

Motivated by the ever-increasing demands for limited communication bandw...

Analog ensemble data assimilation and a method for constructing analogs with variational autoencoders

It is proposed to use analogs of the forecast mean to generate an ensemb...

Sensitivity analysis of chaotic systems using a frequency-domain shadowing approach

We present a frequency-domain method for computing the sensitivities of ...

Code Repositories


MSc Research project (6 months). Data Assimilation using Deep Learning (AEs). Imperial College Machine Learning MSc 2018-19

view repo