# Stationary Distribution for Markov-like system?

Let $$A= \begin{pmatrix} 0 & a_{1,2} & a_{1,3} \\ a_{2,1} & 0 & a_{2,3} \\ a_{3,1} & a_{3,2} & 0 \end{pmatrix},$$

$$B= \begin{pmatrix} 0 & 1 & 1 \\ 1 & 0 & 1 \\ 1 & 1 & 0 \end{pmatrix}$$ with $a_{i,j}>=0$ (no stochastic matrix) and $\pi\in\mathbb{R_{}}_{+}^3$. The system I want to solve is $$\langle\pi,\pi B\rangle=\pi A$$, where $\langle.,.\rangle: \mathbb{R}^3\to\mathbb{R}^3$ is the elementwise multiplication of two vectors. Via programming, I found that with an arbitrary starting vector $\pi_0\in\mathbb{R}_{+}^3$, the following sequence converges to the desired solution $\pi$: $$\pi_{i+1}:=\frac{\pi_{i}A}{\pi_{i}B},$$ where the fraction is again elementwise.

Can anyone help me prove existence and uniqueness of a solution (except the trivial one) to the system via this convergence? It looks a bit like the stationary distribution of irreducible Markov Chains, but it is different.

Also thankful for references to literature or other posts here. Thanks!

Remark: I already asked a quite similar question on Stackexchange Mathematics (https://math.stackexchange.com/questions/709568/how-to-solve-this-system-of-3-equations-with-3-variables) but did not get an answer. I thought MathOverflow would be more suited for the question. Please correct me, if I am wrong.

• Are you interested in nonnegative solutions only? Otherwise, my guess is that the solution is not unique in general. – Federico Poloni Mar 22 '14 at 12:27
• there's a syntax error: $\langle \pi, \pi B\rangle$ is a scalar, whereas $\pi A$ is not... – Suvrit Mar 22 '14 at 12:40
• @Federico: Thx, only nonnegative solutions. – Martin Buber Mar 22 '14 at 12:44
• @Suvrit $\langle \pi, \pi B\rangle$ is not what you think, from the description it should be the Hadamard product. – Federico Poloni Mar 22 '14 at 13:02
• that's terrible notation then I'd say! – Suvrit Mar 22 '14 at 13:04