I am trying to understand the personalised page rank algorithm. I have a graph with multiple types of nodes and I want to calculate how important other nodes are w.r.t these nodes. I am using personalised page rank algorithm for the same.
I followed the notes from this blog https://medium.com/polo-club-of-data-science/pagerank-algorithm-explained-with-examples-a5e25e2594c9
I had two doubts regarding the approach :
a. Since in personalised page-rank, we start from the personalised nodes. Should the initial probability metric should have value only for the personalised nodes? i.e the p matrix should be 1 for the personalisation nodes and 0 for others.
b. Since we will be randomly starting from the personalised nodes, should the value of the probabilities be 1/k instead of 1/n, where k is the number of personalised nodes.
