12,840 Hits in 5.1 sec

Augmenting Physical Models with Deep Networks for Complex Dynamics Forecasting [article]

Yuan Yin, Vincent Le Guen, Jérémie Dona, Emmanuel de Bézenac, Ibrahim Ayed, Nicolas Thome, Patrick Gallinari
2021 arXiv   pre-print
In this work, we introduce the APHYNITY framework, a principled approach for augmenting incomplete physical dynamics described by differential equations with deep data-driven models.  ...  Forecasting complex dynamical phenomena in settings where only partial knowledge of their dynamics is available is a prevalent problem across various scientific fields.  ...  To handle this case, we introduce a principled learning framework to Augment incomplete PHYsical models for ideNtIfying and forecasTing complex dYnamics (APHYNITY).  ... 
arXiv:2010.04456v5 fatcat:bdct4vuywndi7mpnchx732usle

Deep learning for spatio-temporal forecasting – application to solar energy [article]

Vincent Le Guen
2022 arXiv   pre-print
Our second direction is to augment incomplete physical models with deep data-driven networks for accurate forecasting.  ...  For video prediction, we introduce the PhyDNet model that disentangles physical dynamics from residual information necessary for prediction, such as texture or details.  ...  "Augmenting Physical Models with Deep Networks for Complex Dynamics Forecasting", In International Conference on Learning Representations (ICLR 2021, oral presentation), JSTAT 2021.  ... 
arXiv:2205.03571v1 fatcat:dwkprkwf6ncgjcnvkpx3yrdfjm

Data-augmented sequential deep learning for wind power forecasting

Hao Chen, Yngve Birkelund, Qixia Zhang
2021 Energy Conversion and Management  
With excellent automatic pattern recognition and nonlinear mapping ability for big data, deep learning is increasingly employed in wind power forecasting.  ...  The other side unveils that the effect of data augmentations on prediction is intricately varying, but for the proposed model with and without augmentations, all augmentation approaches boost the model  ...  And thanks to the support of UiT Arctic Centre for Sustainable Energy and Dr. Fuqing Yuan and Stian Normann Anfinsen for their comments on the manuscript.  ... 
doi:10.1016/j.enconman.2021.114790 fatcat:ihnafjzmnrasnl36wijzqzzxci

A comparison of data‐driven approaches to build low‐dimensional ocean models

N. Agarwal, D Kondrashov, P. Dueben, E Ryzhov, P. Berloff
2021 Journal of Advances in Modeling Earth Systems  
However, the long timescales of ocean dynamics and the weak influence from the deeper layers of the ocean on the atmosphere for a weather forecast of, say, a couple of days justify the investigation of  ...  Medium-range weather forecast models routinely use computationally expensive ocean general circulation models (OGCMs) that are coupled to the atmosphere model.  ...  , as a less complex network design is required for optimal performance.  ... 
doi:10.1029/2021ms002537 fatcat:wezmw4mxfnd6dcaa2nxpmcql3u

A comparison of data-driven approaches to build low-dimensional ocean models [article]

Niraj Agarwal, Dmitri Kondrashov, Peter Dueben, Evgenii Ryzhov, Pavel Berloff
2021 arXiv   pre-print
Overall, our analysis promotes multi-level LR stochastic models with memory effects, and hybrid models with linear dynamical core augmented by additive stochastic terms learned via deep learning, as a  ...  We found that the multi-level linear stochastic approach performs the best for both short- and long-timescale forecasts.  ...  NA is grateful to the Research Computing Service (RCS) team of Imperial College London for the help and assistance with HPC, and to MPE CDT for providing the financial and technical support for con-  ... 
arXiv:2108.00818v1 fatcat:bdxc5h2hirg4rcdl4oyuyul2ru

Dynamic multistep uncertainty prediction in spatial geometry

Alex Grenyer, Oliver Schwabe, John A. Erkoyuncu, Yifan Zhao
2021 Procedia CIRP  
To ensure continuous forecast accuracy, a conceptual dynamic multistep prediction model is presented applying spatial geometry with long-short term memory (LSTM) neural networks.  ...  Based in MATLAB, this deep learning model predicts uncertainty for the in-service life of a given system.  ...  The authors would like to thank the Engineering and Physical Sciences Research Council (EPSRC), project reference 1944319, and Doctoral Training Partnership (DTP) for funding this research.  ... 
doi:10.1016/j.procir.2021.01.055 fatcat:e4vja52e6zd4xiyntaythqezri

Physics-Guided Deep Learning for Dynamical Systems: A Survey [article]

Rui Wang, Rose Yu
2022 arXiv   pre-print
While deep learning (DL) provides novel alternatives for efficiently recognizing complex patterns and emulating nonlinear dynamics, its predictions do not necessarily obey the governing laws of physical  ...  Modeling complex physical dynamics is a fundamental task in science and engineering. Traditional physics-based models are sample efficient, interpretable but often rely on rigid assumptions.  ...  approximate physical models with deep data-driven networks.  ... 
arXiv:2107.01272v5 fatcat:k6hhdt6csnfebgkzrpuoeqkwzi

Using Neural Architectures to Model Complex Dynamical Systems

Nicholas Gabriel, Neil F. Johnson
2022 Advances in Artificial Intelligence and Machine Learning  
The natural, physical and social worlds abound with feedback processes that make the challenge of modeling the underlying system an extremely complex one.  ...  Modelling complex systems in the physical, biological, and social sciences within the SciML framework requires further considerations.  ...  George Karniadakis for very valuable discussions about PINNs. We are also grateful for funding for this research from the U.S.  ... 
doi:10.54364/aaiml.2022.1124 fatcat:dwakftjcnncjbgamoq3x2qssny

Bounded nonlinear forecasts of partially observed geophysical systems with physics-constrained deep learning [article]

Said Ouala, Steven L. Brunton, Ananda Pascual, Bertrand Chapron, Fabrice Collard, Lucile Gaultier, Ronan Fablet
2022 arXiv   pre-print
The proposed architecture is implemented within a deep learning framework, and its relevance is demonstrated with respect to state-of-the-art schemes for different case-studies representative of geophysical  ...  This is the case in ocean-atmosphere dynamics, for which unknown interior dynamics can affect surface observations.  ...  SLB acknowledges valuable discussions with Alan Kaptanoglu.  ... 
arXiv:2202.05750v2 fatcat:a74lng4dsrenljgfriffhk3vai

AST-GCN: Attribute-Augmented Spatiotemporal Graph Convolutional Network for Traffic Forecasting

Jiawei Zhu, Qiongjie Wang, Chao Tao, Hanhan Deng, Ling Zhao, Haifeng Li
2021 IEEE Access  
We model the external factors as dynamic attributes and static attributes and design an attribute-augmented unit to encode and integrate those factors into the spatiotemporal graph convolution model.  ...  INDEX TERMS Traffic forecasting, graph convolutional network, external factors, spatiotemporal models.  ...  Many deep learning methods have been applied to traffic forecasting, such as deep confidence networks (DBNs) [14] , [15] and stacked autoencoding neural networks (SAEs) [16] .  ... 
doi:10.1109/access.2021.3062114 fatcat:27sptaw2xbgfnb6wt6dj4te2gu

A spatio-temporal LSTM model to forecast across multiple temporal and spatial scales [article]

Yihao Hu, Fearghal O'Donncha, Paulito Palmes, Meredith Burke, Ramon Filgueira, Jon Grant
2021 arXiv   pre-print
This paper presents a novel spatio-temporal LSTM (SPATIAL) architecture for time series forecasting applied to environmental datasets.  ...  Results were compared against RF and XGB baseline models that learned on the temporal signal of each sensor independently by extracting the date-time features together with the past history of data using  ...  ., governing equation variables) by a deep neural network (DNN) and embed the physical law to regularize the network.  ... 
arXiv:2108.11875v1 fatcat:bciqdsock5gqdetlvxxdqnjcn4

Data science and AI in FinTech: An overview [article]

Longbing Cao, Qiang Yang, Philip S. Yu
2021 arXiv   pre-print
blockchain, and the DSAI techniques including complex system methods, quantitative methods, intelligent interactions, recognition and responses, data analytics, deep learning, federated learning, privacy-preserving  ...  processing, augmentation, optimization, and system intelligence enhancement.  ...  Examples are -recurrent neural networks for multivariate timeseries forecasting, temporal and sequential modeling of financial variables, and price or index trend forecasting [51, 26, 52] ; -deep reinforcement  ... 
arXiv:2007.12681v2 fatcat:jntzuwaktjg2hmmjypi5lvyht4

WeatherBench: A benchmark dataset for data‐driven weather forecasting

Stephan Rasp, Peter D. Dueben, Sebastian Scher, Jonathan A. Weyn, Soukayna Mouatadid, Nils Thuerey
2020 Journal of Advances in Modeling Earth Systems  
Further, we provide baseline scores from simple linear regression techniques, deep learning models, as well as purely physical forecasting models.  ...  Data-driven approaches, most prominently deep learning, have become powerful tools for prediction in many domains.  ...  We thank the Copernicus Climate Change Service (C3S) for allowing us to redistribute the data. Peter D.  ... 
doi:10.1029/2020ms002203 fatcat:2esg2brnird5lgzfddvdpkos74

Data Augmentation vs. Equivariant Networks: A Theory of Generalization on Dynamics Forecasting [article]

Rui Wang, Robin Walters, Rose Yu
2022 arXiv   pre-print
Unlike most prior theories for the i.i.d. setting, we focus on non-stationary dynamics forecasting with complex temporal dependencies.  ...  In this work, we derive the generalization bounds for data augmentation and equivariant networks, characterizing their effect on learning in a unified framework.  ...  augmentation and perfectly equivariant networks on modeling imperfectly symmetric dynamics.  ... 
arXiv:2206.09450v1 fatcat:jm3lsuuaxfbdzl24xoklrnilci

Discrepancy Modeling Framework: Learning missing physics, modeling systematic residuals, and disambiguating between deterministic and random effects [article]

Megan R. Ebers, Katherine M. Steele, J. Nathan Kutz
2022 arXiv   pre-print
Physics-based and first-principles models pervade the engineering and physical sciences, allowing for the ability to model the dynamics of complex systems with a prescribed accuracy.  ...  , and (ii) by discovering a model for the missing deterministic physics.  ...  Acknowledgements We are especially grateful to Kadierdan Kaheman and Steven Brunton for discussions related to discrepancy modeling.  ... 
arXiv:2203.05164v1 fatcat:7cj6p4txszhllez6slfczfbrie
« Previous Showing results 1 — 15 out of 12,840 results