Physically interpretable machine learning algorithm on multidimensional non-linear fields

JOURNAL OF COMPUTATIONAL PHYSICS

卷:428

Article

Mouradi, Rem-Sophia^1,2  Goeury, Cedric¹  Thual, Olivier^2,3  Zaoui, Fabrice¹  Tassi, Pablo^1,4 

[1] Natl Lab Hydraul & Environm LNHE, EDF R&D, 6 Quai Watier, F-78400 Chatou, France

[2] French Natl Res Ctr CNRS, European Ctr Res & Adv Training Sci Computat CERF, Climate Environm Coupling & Uncertainties Res Uni, 42 Ave Gaspard Coriolis, F-31820 Toulouse, France

[3] Univ Toulouse, Inst Mecan Fluides Toulouse IMFT, CNRS, Toulouse, France

[4] St Venant Lab Hydraul LHSV, Chatou, France

关键词: Data-Driven Model (DDM);    Proper Orthogonal Decomposition (POD);    Dimensionality Reduction (DM);    Polynomial Chaos Expansion (PCE);    Machine Learning (ML);    Geosciences;   

DOI  :  10.1016/j.jcp.2020.110074

来源: Elsevier

【 摘 要 】

In an ever-increasing interest for Machine Learning (ML) and a favorable data development context, we here propose an original methodology for data-based prediction of two-dimensional physical fields. Polynomial Chaos Expansion (PCE), widely used in the Uncertainty Quantification community (UQ), has long been employed as a robust representation for probabilistic input-to-output mapping. It has been recently tested in a pure ML context, and shown to be as powerful as classical ML techniques for point-wise prediction. Some advantages are inherent to the method, such as its explicitness and adaptability to small training sets, in addition to the associated probabilistic framework. Simultaneously, Dimensionality Reduction (DR) techniques are increasingly used for pattern recognition and data compression and have gained interest due to improved data quality. In this study, the interest of Proper Orthogonal Decomposition (POD) for the construction of a statistical predictive model is demonstrated. Both POD and PCE have amply proved their worth in their respective frameworks. The goal of the present paper was to combine them for a field measurement-based forecasting. The described steps are also useful to analyze the data. Some challenging issues encountered when using multidimensional field measurements are addressed, for example when dealing with few data. The POD-PCE coupling methodology is presented, with particular focus on input data characteristics and training-set choice. A simple methodology for evaluating the importance of each physical parameter is proposed for the PCE model and extended to the POD-PCE coupling. (c) 2020 The Author(s). Published by Elsevier Inc.

【 授权许可】

Free