Data-driven PDE modelling: Trick or treat!

Barczi Jean-François, Melatagia Yonta Paulin, Momo Stéphane Takoudjou, Nzhega Arnauld, Peynaud Emilie, Stinckwich Serge. 2020. Data-driven PDE modelling: Trick or treat!. s.l. : s.n., 1 vidéo (40 min 43 sec)

Résumé : The video presents a multidisciplinary project aimed at testing neural network-based methods for extracting partial differential equations (PDEs) from data to model plant biomass dynamics. PDE models offer an alternative approach to represent and characterize plant growth dynamics in an aggregated manner at the plant or plot scale. However, plants are complex living organisms for which we lack integrated knowledge about spatiotemporal dynamics, though emerging technologies such as LiDAR are providing increasingly diverse datasets. The project explores a modelling approach that combines data and theory in a balanced framework to design PDE models, which initially appear promising. The video describes the multiple challenges encountered when applying these approaches to plant modelling and reveals the methodological obstacles that must be overcome. Finally, the approach is generalized into a workflow called CEDI, designed as a guideline for deriving PDEs from data and structuring the complex task of PDE-based modelling.

Mots-clés libres : Partial differential equations, Neural Network, Physics informed neural networks, Plant modeling, Data-driven theory modelling

Auteurs et affiliations

• Barczi Jean-François, CIRAD-BIOS-UMR AMAP (FRA)
• Melatagia Yonta Paulin, Université de Yaoundé 1 (CMR)
• Momo Stéphane Takoudjou, Université de Yaoundé 1 (CMR)
• Nzhega Arnauld
• Peynaud Emilie, CIRAD-BIOS-UMR AMAP (FRA) ORCID: 0000-0002-2404-8098
• Stinckwich Serge, UNU (CHN)

Source : Cirad-Agritrop (https://agritrop.cirad.fr/613266/)

Métriques

[ Page générée et mise en cache le 2025-09-19 ]