Efficient computation of the Page rank:
We can remove the dangling nodes (Nodes without children) and compute the Page Rank of the remaining nodes. Assuming that the M* matrix doesnt change much we can add these Dangling nodes back with the new probabilities.
This notion is similar to the calculation of the LSI in the document-term matix, where effective computation could have been adding the query to the document matrix (assuming the query is also a document) and computing the LSI but we assume that one query doesn`t alter the pre-computed LSI much hence the query need not be added.
Anuj
