# Why is my PageRank result in Mathematica converging this way?

I am using this technique to find the PR of these 3 pages. It converges to something it shouldn't be. I tried the same iteration in Java and 2 pages come up with 1.333 and the other one as 0.666. Mathematica is calculating something differently. Here's my code:

g = {"Netscape" -> "Amazon", "Netscape" -> "Netscape",
"Amazon" -> "Netscape", "Amazon" -> "Microsoft",
"Microsoft" -> "Amazon"};
pr = PageRankVector[g]
GraphPlot[g,
VertexRenderingFunction -> (Text[
Framed[#2 <> "\n" <>
ToString[pr[[#3]]], {Background -> RGBColor[1, 1, 0.8],
FrameStyle -> RGBColor[0.94, 0.85, 0.36]}], #1] &),
DirectedEdges -> True]

-
You probably have a problem in your Java code or others settings for TeleportProbability. But 1.333 and 0.666 doesn't add up to one, which it should, so I am positive that you have an error in your Java code. – C. E. Jan 27 '14 at 21:44
How do I set Teleport in Mathematica to 0? And in my Java, I am not using any damping at all, so it's the plain PR formula, as described here infolab.stanford.edu/~ullman/mining/websearch.pdf. I am getting the same results as Stanford, so probably because Mathematica is indeed using a teleportation probability. – Ali Gajani Jan 27 '14 at 22:35
You have to look in the documentation.  PageRankVector[g,TeleportProbability->0] – Ajasja Jan 28 '14 at 21:38

## 1 Answer

In the case that this does not closed I'll answer this temporarily (so that there won't be so many unaswerd questions aound)

The problem is TeleportProbability as guessed by @Pickett. To set it to 0 you just have to look in the documentation.

PageRankVector[g,TeleportProbability->0]

-
Unfortunately setting the value to zero will cause The value of option TeleportProbability -> 0 should be a positive machine real number less than 1. and setting it to a small value did not produce the right result for me. I also tried some other things, but alas could not replicate the paper's result. – C. E. Jan 28 '14 at 21:56
@Pickett Should I delete this A then? – Ajasja Jan 29 '14 at 9:00
Probably, the intention was right but in this case TeleportProbability seems to be only a part of the problem. The answer is probably TeleportProbability+normalization+some other difference. – C. E. Jan 29 '14 at 9:03