"I was surprised by the simplicity of the math underlying the google PageRank algorithm, and the ease with which it seemed to be efficiently implementable" It'

"I was surprised by the simplicity of the math underlying the google PageRank algorithm, and the ease with which it seemed to be efficiently implementable" It's actually not all that easy to implement this stuff efficiently when n grows to any sort of useful size. E.g. the article uses an n x n adjacency matrix, so if we take n = 1 million pages, we need 4TB just to represent it, assuming no overhead. (Adjacency lists help here.) See http://dbpubs.stanford.edu:8090/pub/showDoc.Fulltext?lang=en&doc=2003-16&format=pdf&compression=&name=2003-16.pdf http://dbpubs.stanford.edu:8090/pub/showDoc.Fulltext?lang=en.... (I hacked together a script to compute this on MediaWiki wikis a while ago, by examining the DB directly -- http://collison.ie/code/pagerank.rb http://collison.ie/code/pagerank.rb)