It does deal with dangling nodes and makes sure that the transition matrix is irreducible, but still the ARPACK eigenvector solver fails to converge every now and then and it's pretty hard to figure out what's going on because ARPACK is a black box for us (igraph just multiplies matrices whenever ARPACK asks igraph to do so). We plan to switch to PRPACK (see http://github.com/DavidKurokawa/prpack) soon and this will hopefully resolve the problems with personalized PageRank.
It does deal with dangling nodes and makes sure that the transition matrix is irreducible, but still the ARPACK eigenvector solver fails to converge every now and then and it's pretty hard to figure out what's going on because ARPACK is a black box for us (igraph just multiplies matrices whenever ARPACK asks igraph to do so). We plan to switch to PRPACK (see http:// github. com/DavidKuroka wa/prpack) soon and this will hopefully resolve the problems with personalized PageRank.