Operator Inference needs a robust implementation that goes beyond the Linear least squares approach with simple regularizers that they implement here https://github.com/Willcox-Research-Group/rom-operator-inference-Python3 .
In principle there are two improvements off the top of my head.
- Gradient computation from snapshots need to be carried out with robust Fornberg methods : https://web.njit.edu/~jiang/math712/fornberg.pdf
- Least squares approach assumes linear and quadratic forms for the operator A. This has to be generalized by leveraging Optim.jl
Operator Inference needs a robust implementation that goes beyond the Linear least squares approach with simple regularizers that they implement here https://github.com/Willcox-Research-Group/rom-operator-inference-Python3 .
In principle there are two improvements off the top of my head.