0
$\begingroup$

Given a random $d$-regular graph on $n$ nodes, what is the expected number of common neighbors between two nodes?

I don't know if it is as simple as just assuming that each neighbor of the first node has a $\frac{d}{n}$ probability of being a neighbor of the second, as the set of $d$-regular graphs on $n$ nodes is difficult to construct.

$\endgroup$
2
  • $\begingroup$ Not quite as simple if you are looking for the exact value. If $d=1$, you will certainly get a wrong answer this way and there is no reason to believe that other values of $d$ are any better. $\endgroup$
    – fedja
    Sep 10, 2018 at 20:21
  • $\begingroup$ @fedja Well if $d=1$, then the expected value is 0. Some experimental results using networkx's random_regular_graph (which, if I'm reading the API right, seems to be good if $d<n^{\frac{1}{3}}$) gives E(n=1000, d=10) as approx .09. For E(n=1000, d=8) I got about .055. $\endgroup$ Sep 10, 2018 at 21:13

1 Answer 1

2
$\begingroup$

Yep, it is $\frac{d(d-1)}{n-1}$ in general. Fix a $d$-regular graph and average over the action of the permutation group on the vertices. We are interested in the event $S(x,y,z)=$"$x$ is connected to $y$ and $z$". If $E$ is the expected number of neighbors, then $\sum_{x,y,z}P(S(x,y,z))=n(n-1)E$, summing over $x$ first. On the other hand, if we sum over $y,z$ first, the same sum is evaluated as $nd(d-1)$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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