Multiresolution convolutional autoencoders

We propose a multi-resolution convolutional autoencoder (MrCAE) architecture that integrates and leverages three highly successful mathematical architectures: (i) multigrid methods, (ii) convolutional autoencoders and (iii) transfer learning. The method provides an adaptive, hierarchical architectur...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of computational physics Jg. 474; S. 111801
Hauptverfasser: Liu, Yuying, Ponce, Colin, Brunton, Steven L., Kutz, J. Nathan
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Inc 01.02.2023
Schlagworte:
Convolutional autoencoder
Model scaling
Multi-scale dynamics
Multigrid
Multiresolution analysis
Transfer learning
Model scaling
Multi-scale dynamics
Multigrid
Transfer learning
Convolutional autoencoder
Multiresolution analysis
ISSN:0021-9991, 1090-2716
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:We propose a multi-resolution convolutional autoencoder (MrCAE) architecture that integrates and leverages three highly successful mathematical architectures: (i) multigrid methods, (ii) convolutional autoencoders and (iii) transfer learning. The method provides an adaptive, hierarchical architecture that capitalizes on a progressive training approach for multiscale spatio-temporal data. This framework allows for inputs across multiple scales: starting from a compact (small number of weights) network architecture and low-resolution data, our network progressively deepens and widens itself in a principled manner to encode new information in the higher resolution data based on its current performance of reconstruction. Basic transfer learning techniques are applied to ensure information learned from previous training steps can be rapidly transferred to the larger network. As a result, the network can dynamically capture different scaled features at different depths of the network. The performance gains of this adaptive multiscale architecture are illustrated through a sequence of numerical experiments on synthetic examples and real-world spatial-temporal data. •This work develops a multiresolution convolutional autoencoder architecture for characterizing spatially multi-scale dynamical systems.•Transfer learning is leveraged to progressively grow the size of the network, allowing in-puts across multiple scales.•Adaptive convolutional filters are used to exploit the multi-scale features of the data, resembling mesh refinement.•The flexible workflow produces highly compact architectures and interpretable encodings, giving rise to an in-depth understanding over network architectures.
ISSN:0021-9991
1090-2716
DOI:10.1016/j.jcp.2022.111801