Benchmark current performance on public datasets

[pankajgupta]

Use datasets on http://snap.stanford.edu/data/index.html and benchmark performance of a couple of algorithms for very big graphs, such as (1) Global pagerank, (2) Personalized pagerank for every node in the graph.

[szymonm]

Hi Pankaj, this issue is quite old, is it still available?

I see your global pagerank implementation in algorithms package. I guess I should start with coding personalized pagerank. Am I correct, that the only difference between personalized and the global one is the non-uniform probabilities vector for jumps? In your implementation it means a an array of dampingAmount instead of one value. This should be passed as a parameter to the algorithm. Am I right?

So shouldn't I generalize the global one? For example by adding optional parameter Function1[Int, Double](default _ => 1) to the PageRankParams?

[pankajgupta]

Yes, this is still valid. Personalized pagerank already exists. See method calculatePersonalizedReputation(…) in https://github.com/twitter/cassovary/blob/master/src/main/scala/com/twitter/cassovary/graph/GraphUtils.scala#L136

[szymonm]

This should be closed.