top of page

DATA-DRIVEN MODELING

Data-Driven: Projects

ROBUST PDE IDENTIFICATION FROM NOISY DATA

Given the vast development of data acquisition techniques, data-driven modeling receives more attention in recent years. It is very challenging to identify a stable model from noisy data, especially when it contains unbounded operators, such as differential operators. We proposed an effective smoothing technique tailored for this issue. Moreover, we employed the subspace pursuit algorithm to generate candidate models with various options of sparsity, and two selection paradigms are developed to filter the model best approximating the underlying PDE.

[He et al. 2020, Submitted, arXiv ]

PDE.png
bottom of page