Comments on: The PageRank distribution for the web http://michaelnielsen.org/blog/the-pagerank-distribution-for-the-web/ Mon, 07 May 2012 21:35:26 +0000 hourly 1 http://wordpress.org/?v=3.3 By: Tecnologia de Internet (Locaweb) » Blog Archive » RailsConf’09: painel do primeiro dia http://michaelnielsen.org/blog/the-pagerank-distribution-for-the-web/comment-page-1/#comment-21850 Tecnologia de Internet (Locaweb) » Blog Archive » RailsConf’09: painel do primeiro dia Wed, 06 May 2009 04:24:27 +0000 http://michaelnielsen.org/blog/?p=516#comment-21850 [...] De tarde, tivemos a apresentação do Ilya Grigorik (AideRSS Inc.), com o título: “Building a Mini-Google: High-Performance Computing in Ruby“, que falou sobre um algoritmo de classificação e sua implementação em Ruby, a apresentação foi baseada a partir de uma sequência de posts de Michel Nielsen. [...] [...] De tarde, tivemos a apresentação do Ilya Grigorik (AideRSS Inc.), com o título: “Building a Mini-Google: High-Performance Computing in Ruby“, que falou sobre um algoritmo de classificação e sua implementação em Ruby, a apresentação foi baseada a partir de uma sequência de posts de Michel Nielsen. [...]

]]> By: Michael Nielsen http://michaelnielsen.org/blog/the-pagerank-distribution-for-the-web/comment-page-1/#comment-16268 Michael Nielsen Mon, 15 Dec 2008 03:10:58 +0000 http://michaelnielsen.org/blog/?p=516#comment-16268 No, I don't use a mac. In any case, I'm not in any hurry. I intend to work my way through the documentation for the Python scientific libraries, and I'll come to it. I've read a bunch of Barabasi's papers, and by other people (Watts, Newman etc) working on power laws and graphs etc. It's certainly interesting stuff! I haven't yet read "Linked", although it's been sitting on my bookcase for a few years, and I should probably take it down. No, I don’t use a mac. In any case, I’m not in any hurry. I intend to work my way through the documentation for the Python scientific libraries, and I’ll come to it.

I’ve read a bunch of Barabasi’s papers, and by other people (Watts, Newman etc) working on power laws and graphs etc. It’s certainly interesting stuff! I haven’t yet read “Linked”, although it’s been sitting on my bookcase for a few years, and I should probably take it down.

]]> By: Artur Adib http://michaelnielsen.org/blog/the-pagerank-distribution-for-the-web/comment-page-1/#comment-16266 Artur Adib Mon, 15 Dec 2008 03:02:16 +0000 http://michaelnielsen.org/blog/?p=516#comment-16266 Do you use a Mac? If so, check out this nice little plotting program (which I actually use for research): http://plot.micw.eu/ You seem to be stepping into "Barabasi grounds", which is pretty exciting. Have you read his book "Linked"? It's a gentle introduction to how power law behavior emerges in graph theory. I enjoyed it. -A. Do you use a Mac? If so, check out this nice little plotting program (which I actually use for research):

http://plot.micw.eu/

You seem to be stepping into “Barabasi grounds”, which is pretty exciting. Have you read his book “Linked”? It’s a gentle introduction to how power law behavior emerges in graph theory. I enjoyed it.

-A.

]]> By: Michael Nielsen http://michaelnielsen.org/blog/the-pagerank-distribution-for-the-web/comment-page-1/#comment-16262 Michael Nielsen Mon, 15 Dec 2008 01:24:48 +0000 http://michaelnielsen.org/blog/?p=516#comment-16262 Artur - I'm brand new to Python, and spent about 20 minutes twiddling with libraries trying to get log-log before I posted. I'll get it later, once I've read through all the documentation for the numeric library. Thanks for the link to the paper, which I hadn't seen. Artur – I’m brand new to Python, and spent about 20 minutes twiddling with libraries trying to get log-log before I posted. I’ll get it later, once I’ve read through all the documentation for the numeric library. Thanks for the link to the paper, which I hadn’t seen.

]]> By: Michael Nielsen http://michaelnielsen.org/blog/the-pagerank-distribution-for-the-web/comment-page-1/#comment-16260 Michael Nielsen Mon, 15 Dec 2008 01:18:14 +0000 http://michaelnielsen.org/blog/?p=516#comment-16260 Oz - PageRank isn't generally proportional to the indegree. Getting a measure that isn't directly related to indegree was part of Page and Brin's motivation, because indegree is so easy to game by spammers. With that said, some people have noticed a reasonable correlation between the two measures in many instances. Oz – PageRank isn’t generally proportional to the indegree. Getting a measure that isn’t directly related to indegree was part of Page and Brin’s motivation, because indegree is so easy to game by spammers. With that said, some people have noticed a reasonable correlation between the two measures in many instances.

]]> By: justkeeper http://michaelnielsen.org/blog/the-pagerank-distribution-for-the-web/comment-page-1/#comment-16259 justkeeper Mon, 15 Dec 2008 00:34:43 +0000 http://michaelnielsen.org/blog/?p=516#comment-16259 You definitely should have a look at this article,http://arxivblog.com/?p=558 , which is very interesting. You definitely should have a look at this article,http://arxivblog.com/?p=558 , which is very interesting.

]]> By: oz http://michaelnielsen.org/blog/the-pagerank-distribution-for-the-web/comment-page-1/#comment-16257 oz Mon, 15 Dec 2008 00:18:37 +0000 http://michaelnielsen.org/blog/?p=516#comment-16257 Isn't the page-rank of a page simply the fraction of time a random-walker spends on this page? Assuming this, if the network was undirected, then the page-rank is exactly proportional to the degree of the node. Thus if the degree is taken from a power-law distribution, it immediately implies that the page-rank follows a power-law with the same exponent. For a directed graph i guess there are deviations from this, but my intuition is that if edges are drawn randomly it wouldn't change much thus it's not surprising that you still get a power-law distribution. It's interesting and easy to check if the exponent of the page-rank dist. is the same as for the in-degree dist. Isn’t the page-rank of a page simply the fraction
of time a random-walker spends on this page?

Assuming this, if the network was undirected, then the page-rank is exactly proportional to the degree of the node. Thus if the degree is taken from a power-law distribution, it immediately implies that the page-rank follows a power-law with the same exponent.
For a directed graph i guess there are deviations from this, but my intuition is that if edges are drawn randomly it wouldn’t change much thus it’s not surprising that you still get a power-law distribution. It’s interesting and easy to check if the exponent of the page-rank dist. is the same as for the in-degree dist.

]]> By: Artur Adib http://michaelnielsen.org/blog/the-pagerank-distribution-for-the-web/comment-page-1/#comment-16225 Artur Adib Sat, 13 Dec 2008 23:19:22 +0000 http://michaelnielsen.org/blog/?p=516#comment-16225 I know this is trivial, but how about using a log-log scale for your plots? We can hardly see the data in this linear scale. Perhaps this will also reveal your power-law. You might also enjoy taking a look at what folks from my alma mater did on the subject: "Using PageRank to Characterize Web Structure" http://www.internetmathematics.org/volumes/3/1/Pandurangan.pdf I'm really enjoying your lectures though. Keep up the good work. -Artur http://arturadib.blogspot.com I know this is trivial, but how about using a log-log scale for your plots? We can hardly see the data in this linear scale. Perhaps this will also reveal your power-law.

You might also enjoy taking a look at what folks from my alma mater did on the subject:

“Using PageRank to Characterize Web Structure”

http://www.internetmathematics.org/volumes/3/1/Pandurangan.pdf

I’m really enjoying your lectures though. Keep up the good work.

-Artur
http://arturadib.blogspot.com

]]>