If $p(i)$ is the probability of state $i$ with $p(i)>0$ and $p_{ij}=p(j.i^{-1})$ are transition probabilities where product and inverse refer to group operations, how can I show that the transition probability matrix is doubly stochastic?


Note that to show that the matrix $P$, with coordinates $p_{ij}$, is doubly stochastic, it is enough to show that the values of each rows and each column sum to $1$.

For this, observe that $\sum\limits_i p(i) = 1$.

Now consider the $j'$th column, the sum of of elements in this column is given by $$\sum\limits_i p_{ij} = \sum\limits_i p(j \cdot i^{-1}) = \sum\limits_{j^{-1}\cdot i^{-1}|i \in G} p(i) = 1,$$ where the second equality is just a change of variables.

The same logic also holds for the sum of rows, which gives the desired result.

  • $\begingroup$ I'm not sure I understand the second equality. Could you please break it down or explain for me? Should the range of the sum be $j^{-1}\cdot i|i \in G$? How does it sum to 1? I think I'm quite naive! $\endgroup$ – Blain Waan Oct 17 '18 at 21:21
  • 1
    $\begingroup$ The main idea is that to go over all elements in the group of the type $j \cdot i^{-1}$ (where $j$ is fixed and $i$ is every possible element in $G$), is exactly like simply going over all elements in the group. To see this, for $g \in G$, just take $i = g^{-1} \cdot j$, so that $j\cdot i^{-1} = g$. $\endgroup$ – Cain Oct 17 '18 at 21:29

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.