This is a read only copy of the old FEniCS QA forum. Please visit the new QA forum to ask questions

Computing approximation of Schur complement

0 votes

Hi, I would like to compute an approximation of Schur complement or even the exact Schur complement in FEniCS for saddle-point matrix (of size (n+m)x(n+m))

[Kaa, Kab]
[Kba,  0 ],

i.e. a matrix S = - Kba * inverse(Kaa) * Kab of size (mxm).

I found a simple guide for PETSc here but I cannot transfare it to FEniCS language (python version). Can you help me, please? Any other suggestions how to do the computation are welcomed.

asked Sep 9, 2015 by Ianx86 FEniCS User (1,050 points)
edited Sep 11, 2015 by Ianx86

The guide you link is for computing the exact Schur complement. Is that what you want?

commented Sep 11, 2015 by Magne Nordaas FEniCS Expert (13,820 points)

For this moment, yes. In the future I will go for some problem-motivated approximations.

commented Sep 11, 2015 by Ianx86 FEniCS User (1,050 points)
...