Re: [igraph] Weighted PageRank

[Tamas Nepusz]

Hello,

> However in my implementation a certain link has a probability of existence in
> a specific time period e.g. it is present in my experimentions with
> probability 0.6 or so. Is there a way to apply such a modificationn i.e.
> PR(A) = 1-d +d( (PR(B)/outDeg(B)*(probability of (B->A) to exist) + ..... )
Does it apply to *all* of the links in your network? In that case, nothing will
change because the probabilities of traversing an edge are multiplied by 0.6,
but the outdegree is also multiplied by 0.6, and these modifications cancel
each other out.

> For the weighted Pagerank implementation i assume that the weights on the
> edges depict the probability for going from i->j from example, i.e. not 1/N
> but with 0.6 probability, am i right?
More or less - the weights are *proportional* to the probabilities. So, if you
have a node with two outgoing edges, both of which have weight 0.6, it means
that both of them will be followed by the _same_ probability when your random
walker is in the origin node of these edges and has decided to take a "regular"
step and not a teleportation step. So, the probability of following one of
these edges is therefore 0.5. It will be 0.5 no matter how large or small the
weights are as long as they are equal and nonzero.

Best,
T.