Closed KikeM closed 3 years ago
Okay, so, the linear operator's offline step is actually not that painful at all. What is actually painful is the construction of the collateral basis for the nonlinear operator.
The main difficulty, in comparison with starting fresh for the reduced basis, is the fact that I need to build the reduced mesh too.
I think I ought to simply create a method that allows me to load the basis (they are numpy arrays), go through the reduced mesh computation, and then evaluate and else. The key is that this should be done per run, not inside a loop in the code.
From the spectrum, which I should save, I can determine how many operators I want to select for the sensitivity analysis.
This is useful for the sensitivity analysis.