A priori physical information to aid generalization capabilities of neural networks for hydraulic modeling
The application of Neural Networks to river hydraulics and flood mapping is fledgling, despite the field suffering from data scarcity, a challenge for machine learning techniques. Consequently, many purely data-driven Neural Networks have shown limited capabilities when tasked with predicting new sc...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2025-01-01
|
| Series: | Frontiers in Complex Systems |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fcpxs.2024.1508091/full |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841558752796344320 |
|---|---|
| author | Gianmarco Guglielmo Andrea Montessori Jean-Michel Tucny Michele La Rocca Pietro Prestininzi |
| author_facet | Gianmarco Guglielmo Andrea Montessori Jean-Michel Tucny Michele La Rocca Pietro Prestininzi |
| author_sort | Gianmarco Guglielmo |
| collection | DOAJ |
| description | The application of Neural Networks to river hydraulics and flood mapping is fledgling, despite the field suffering from data scarcity, a challenge for machine learning techniques. Consequently, many purely data-driven Neural Networks have shown limited capabilities when tasked with predicting new scenarios. In this work, we propose introducing physical information into the training phase in the form of a regularization term. Whereas this idea is formally borrowed from Physics-Informed Neural Networks, the proposed methodology does not necessarily resort to PDEs, making it suitable for scenarios with significant epistemic uncertainties, such as river hydraulics. The method enriches the information content of the dataset and appears highly versatile. It shows improved predictive capabilities for a highly controllable, synthetic hydraulic problem, even when extrapolating beyond the boundaries of the training dataset and in data-scarce scenarios. Therefore, our study lays the groundwork for future employment on real datasets from complex applications. |
| format | Article |
| id | doaj-art-2405a0dd5a3e480d816c5e63d989ec22 |
| institution | Kabale University |
| issn | 2813-6187 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Complex Systems |
| spelling | doaj-art-2405a0dd5a3e480d816c5e63d989ec222025-01-06T05:13:19ZengFrontiers Media S.A.Frontiers in Complex Systems2813-61872025-01-01210.3389/fcpxs.2024.15080911508091A priori physical information to aid generalization capabilities of neural networks for hydraulic modelingGianmarco GuglielmoAndrea MontessoriJean-Michel TucnyMichele La RoccaPietro PrestininziThe application of Neural Networks to river hydraulics and flood mapping is fledgling, despite the field suffering from data scarcity, a challenge for machine learning techniques. Consequently, many purely data-driven Neural Networks have shown limited capabilities when tasked with predicting new scenarios. In this work, we propose introducing physical information into the training phase in the form of a regularization term. Whereas this idea is formally borrowed from Physics-Informed Neural Networks, the proposed methodology does not necessarily resort to PDEs, making it suitable for scenarios with significant epistemic uncertainties, such as river hydraulics. The method enriches the information content of the dataset and appears highly versatile. It shows improved predictive capabilities for a highly controllable, synthetic hydraulic problem, even when extrapolating beyond the boundaries of the training dataset and in data-scarce scenarios. Therefore, our study lays the groundwork for future employment on real datasets from complex applications.https://www.frontiersin.org/articles/10.3389/fcpxs.2024.1508091/fullneural networksphysical training strategiesriver hydraulicshydraulic modelinggeneralization |
| spellingShingle | Gianmarco Guglielmo Andrea Montessori Jean-Michel Tucny Michele La Rocca Pietro Prestininzi A priori physical information to aid generalization capabilities of neural networks for hydraulic modeling Frontiers in Complex Systems neural networks physical training strategies river hydraulics hydraulic modeling generalization |
| title | A priori physical information to aid generalization capabilities of neural networks for hydraulic modeling |
| title_full | A priori physical information to aid generalization capabilities of neural networks for hydraulic modeling |
| title_fullStr | A priori physical information to aid generalization capabilities of neural networks for hydraulic modeling |
| title_full_unstemmed | A priori physical information to aid generalization capabilities of neural networks for hydraulic modeling |
| title_short | A priori physical information to aid generalization capabilities of neural networks for hydraulic modeling |
| title_sort | priori physical information to aid generalization capabilities of neural networks for hydraulic modeling |
| topic | neural networks physical training strategies river hydraulics hydraulic modeling generalization |
| url | https://www.frontiersin.org/articles/10.3389/fcpxs.2024.1508091/full |
| work_keys_str_mv | AT gianmarcoguglielmo aprioriphysicalinformationtoaidgeneralizationcapabilitiesofneuralnetworksforhydraulicmodeling AT andreamontessori aprioriphysicalinformationtoaidgeneralizationcapabilitiesofneuralnetworksforhydraulicmodeling AT jeanmicheltucny aprioriphysicalinformationtoaidgeneralizationcapabilitiesofneuralnetworksforhydraulicmodeling AT michelelarocca aprioriphysicalinformationtoaidgeneralizationcapabilitiesofneuralnetworksforhydraulicmodeling AT pietroprestininzi aprioriphysicalinformationtoaidgeneralizationcapabilitiesofneuralnetworksforhydraulicmodeling AT gianmarcoguglielmo prioriphysicalinformationtoaidgeneralizationcapabilitiesofneuralnetworksforhydraulicmodeling AT andreamontessori prioriphysicalinformationtoaidgeneralizationcapabilitiesofneuralnetworksforhydraulicmodeling AT jeanmicheltucny prioriphysicalinformationtoaidgeneralizationcapabilitiesofneuralnetworksforhydraulicmodeling AT michelelarocca prioriphysicalinformationtoaidgeneralizationcapabilitiesofneuralnetworksforhydraulicmodeling AT pietroprestininzi prioriphysicalinformationtoaidgeneralizationcapabilitiesofneuralnetworksforhydraulicmodeling |