Hallo,
I want to calculate a 2D nonlinear Maxwell equation. Therefor I have a form
r(w, u,v)=nu(grad w)grad u * grad v*dx
with v test function, u trial function, nu(grad w) nonlinear function which gives a number (goes to 1 dim. space).
The Maxwell equation is than given as r(u,u,v)=l(v). l(v) a linear form.
Using FEM gives A(u)\zeta=b with u=\sum \zeta_i \phi_i (\phi_i basis functions of trial space).
Now I want to use DEIM (model order reduction) to make this fast.
The problem is:
I have an approx. of u, call it w -> now I want to evaluate A(w)_ij, i.e. the i,j-th component of A(w). This could be done by assemble(r(w,u,v)) u,v, test/trial space, w chosen, and read out the needed ones. But I need only a few components of the matrix and want to make it faster by calculating only this ones. Is this possible with fenics/dolphin?
