Exploration of the Google PageRank Algorithm

Re:Exploration of the Google PageRank Algorithm by wfunction on Thursday May 02, 2013 @05:41AM (#5643433)

In a real application, you don't even know the transition probabilities to begin with, so actually performing the computation is the least of your worries -- the first problem is coming up with numbers, which wasn't mentioned. :)

So I feel that if the goal is to give a conceptual understanding, linear algebra is overkill. If the goal is to show how it's done in practice then this isn't terribly useful since it doesn't even show the tip of the iceberg!

[ Parent | on HN ]

Re:Exploration of the Google PageRank Algorithm by bflbfl on Thursday May 02, 2013 @08:19PM (#5648118)

Hey wfunction

I believe we know exactly what the matrix is, based on the hyperlink structure itself. It's updated and displayed on the vis here based on that, too.

btw, David Austin's article on PageRank is highly recommended : http://www.ams.org/samplings/feature-column/fcarc-pagerank

[ Parent | on HN ]

Re:Exploration of the Google PageRank Algorithm by mullr on Thursday May 02, 2013 @05:58AM (#5643486)

Fair enough. I'm thinking in particular of a paper I read that used pagerank for word sense disambiguation [1]. It's actually variant of pagerank, so the high level description in terms of probabilities (which I read several of) didn't take me very far. In order to come to grips with the technique I ended up actually implementing it and trying it on some toy datasets. This definitely would have helped me in that case.

[1] http://www.aclweb.org/anthology/E/E09/E09-1005.pdf It's interesting research, but it unfortunately turns out that even the best results in word sense disambiguation aren't good enough for a lot of applications. I wish I had 2 years to do nothing but play around in this field.

[ Parent | on HN ]