0
$\begingroup$

I am searching a solution method for the following equation system of equation systems:

Let $A, B \in \mathbb{R}^{n \times n}$ be s.p.d. Matrices and $O$ be the zero matrix of the same size. Further let $f\in\mathbb{R}^n$ be given vector and $0$ the zero vector. I am searching for a numerical method to find the solution vectors $x_1, x_2 \in\mathbb{R}^n$ such that

\begin{equation} \begin{pmatrix} A & B \\ O & A \end{pmatrix} \begin{pmatrix} x_1 \\ x_2 \end{pmatrix} = \begin{pmatrix} f \\ 0 \end{pmatrix} \end{equation}

I tried to solve this in Python with (scipy.linalgscipy.linalg: lu_factor, lu_solve). The problem, I only get the trivial solution $x_2 = 0$. Is there a way to solve the equation system of equation systems for non-trivial solutions?

$\endgroup$
  • $\begingroup$ That looks like a saddle point system / Lagrange multiplier system gone awry $[\mathbf A \mathbf B; \mathbf B^T \mathbf 0]$. I'd double check your reference material / formulation, then ask a followup. There are special methods (both iterative and direct) for those kinds of systems. $\endgroup$ – rchilton1980 Nov 6 at 20:34
4
$\begingroup$

There is no way because there is no solution to the given system with $x_2 \neq 0$. This is because the second block of the equation system reads $Ax_2 = 0$, which has no non-trivial solutions because $A$ is SPD (Otherwise $x_2$ would be an eigenvector of $A$ corresponding to a zero eigenvalue).

$\endgroup$
  • $\begingroup$ Oh you are totally correct. I overlooked the fact of spd. Maybe something in my calculation is not correct. I will write again! $\endgroup$ – Kerem Nov 6 at 11:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.