3
$\begingroup$

Let $Q$ be a $n\times n$ reducible stochastic matrix. Let $J$ be such that $[J]_{ij}={1 \over n}$. Now for a small positive constant $\alpha\in [0,1]$, consider the matrix

$$\tilde{Q}\,=\,(1-\alpha)Q+\alpha J$$

Though $Q$ is reducible meaning it has a eigenspace of dimension more than $1$ corresponding to the Perron root, $\tilde{Q}$ is irreducible guaranteeing a unique Perron vector. So this transition from reducibility of $Q$ to irreducibility of $\tilde{Q}$ is a function of $\alpha$. I was curious if we can study the Perron vector as a function of $\alpha$. Has there been any study in literature on this?

PS: I know that this is referred to as the teleportation probability in computer science (for example, the famous Pagerank algorithm uses this). Is there any other application area where this is used frequently?

Improve this question
$\endgroup$
1
  • $\begingroup$ If you were curious about the behavior of the Perron root in the limit that $\alpha \to 0$, I recently answered a similar question here on math.SE. I show that this limit exists and how to calculate it. I was hoping someone else could review the proof and see if there are any holes in it, if you are interested. $\endgroup$
    – sasquires
    Commented Jul 28, 2020 at 5:31

1 Answer 1

Reset to default
4
$\begingroup$

So this transition from reducibility of Q to irreducibility of $\tilde Q$ is a function of α.

It is easy to show that for any $\alpha > 0$ $\tilde Q$ will be irreducible. See Lemma 3.1 from A note on perturbations of stochastic matrices by Bertram and Wolfgang).

Applying the Theorem 3.2 from the same note, we obtain that Perron vector $P(\alpha)$ is a analytic (thus smooth, and continous) function.

(Actually, it should be a comment. But, unfortunately, I cannot add any comments)

Improve this answer
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .