| Open achives Hal IMT Atlantique |
Learning Variational Data Assimilation Models and Solvers
Abstract : This paper addresses variational data assimilation from a learning point of view. Data assimilation aims to reconstruct the time evolution of some state given a series of observations, possibly noisy and irregularly-sampled. Using automatic differentiation tools embedded in deep learning frameworks, we introduce end-to-end neural network architectures for data assimilation. It comprises two key components: a variational model and a gradient-based solver both implemented as neural networks. A key feature of the proposed end-to-end learning architecture is that we may train the NN models using both supervised and unsupervised strategies. Our numerical experiments on Lorenz-63 and Lorenz-96 systems report significant gain w.r.t. a classic gradient-based minimization of the variational cost both in terms of reconstruction performance and optimization complexity. Intriguingly, we also show that the variational models issued from the true Lorenz-63 and Lorenz-96 ODE representations may not lead to the best reconstruction performance. We believe these results may open new research avenues for the specification of assimilation models in geoscience.
Cited literature [35 references]
https://hal-imt-atlantique.archives-ouvertes.fr/hal-02906798
Contributor : Ronan Fablet <>
Submitted on : Saturday, July 25, 2020 - 6:21:19 PM
Last modification on : Wednesday, August 5, 2020 - 3:43:07 AM
Contributor : Ronan Fablet <>
Submitted on : Saturday, July 25, 2020 - 6:21:19 PM
Last modification on : Wednesday, August 5, 2020 - 3:43:07 AM
File
learning4DVar (16).pdf
Files produced by the author(s)