fdaPDE is a library for physics-informed statistical learning of spatial and functional data, at the intersection between statistics and numerical analysis, for data located over complex multidimensional domains. The use of Partial Differential Equations (PDEs) permits to borrow informations coming from the physics of the problem under study into the statistical modeling, making fdaPDE an extremly flexible tool for the analysis of complex data.
fdaPDE offers a wide range of modeling capabilities, such as regression, non-parametric density estimation, functional data analysis and more, for data located over a spatial domain, possibly evolving in time.
fdaPDE is a C++ library which interfaces with R, one of the most popular language for data analysis, and runs on all major platforms (Linux, Mac, Windows). To get started, see the documentation site.
Currently a new version of the library is under development, the documentation site refers to such development stage
The fdaPDE team will present a short course on physics-informed statistical learning at the International Workshop on Statistical Modeling (IWSM25), taking place in Limerick, Ireland, on July 13, 2025. For details on the topics covered, visit this link.
If you use fdaPDE for any scientific publication, please cite our project with:
Arnone, E., Clemente, A., Sangalli, L.M., Lila, E., Ramsay, J., and Formaggia, L. (2023),
fdaPDE: Physics-Informed Spatial and Functional Data Analysis, R package available from CRAN
https://cran.r-project.org/package=fdaPDE
