Paolo Castagna wrote: > Sebastian Schelter wrote: >> I guess that summing up and redistributing the pagerank of dangling >> vertices can also be done without an extra superstep in an aggregator. > > Yeah! Why didn't I think about that? > Thanks, great suggestion. > > I am going to give this a go, at first without extending RandomWalkVertex > since I want to see how it might work. > This would also inform design and improvements of the current > RandomWalkVertex.
You can find a 'proper' ;-) implementation of PageRank with dangling nodes support, sum at the end of all PageRank values equals to 1.00 (since it's a probability distribution) and PageRank values validated against a third party implementation (i.e. JUNG), here [1]. I have not managed to do it as Sebastian suggested yet. Paolo [1] https://github.com/castagna/jena-grande/blob/3c2d9f85bacb737acd575e3e287dc0fcc6bd96b9/src/main/java/org/apache/jena/grande/giraph/pagerank/PageRankVertex.java