Page Rank is not a particularly impressive algorithm. The technical leap that Google made was the idea of having outgoing links be a proxy for authority, and it was an innovative and useful observation. Page Rank is about the second-simplest implementation of that concept (the first-simplest would not have the dampening factor). It's just a glorified eigenvector problem. Page Rank is not impressive for being a significant technical breakthrough. In fact, I think Page Rank is more impressive for not being a technical breakthrough. Google came up with a single good idea, and even a very very simple application of that idea had significant practical benefit.
Actually, no PageRank wasn't even the first web search algorithm to do that. HITS, by Jon Kleinberg was the first rigorous formulation of that notion (which is mentioned in the article) and was cited in the original PageRank paper. It was widely known at the time that it was a concept imported from library science for correlation in co-citation patterns.
Uhm, Jon Kleinberg is an academic (one of computer science's finest) and was clearly cited in the PageRank paper. He's authored seminal publications across several sub-fields of computer science, he's in a named chair in an Ivy League university, and laid the foundations for modern web search. I'd say he's been pretty successful for a guy in his 30s.
I may have misunderstood you, but isn't a node with a lot of outgoing links said to be a hub and a node with a lot of incoming links said to be an authority ?
Yeah, sorry, my mistake. I think "a page where lots of other pages have outgoing links pointed at it" was what ran through my head... but that's what "lots of incoming links" means. Too late to edit!
Kalman didn't invent the Kalman filter. He learned about the discrete equivalent from Stratonovich and adapted it to the continuous case, and that was the name that stuck. But every useful Kalman filter is actually a Stratonovich filter, because they're all digital.
I guess having a short name makes you more likely to become immortal.
I took a class with Jon Kleinberg called Networks co-taught by an economics professor. Introduced me to power laws, Nash equili riums, and six degrees of separation(or rather the validity thereof) Brilliant class. Brilliant guy.
Google's search quality very likely doesn't have very much to do with the PageRank algorithm as such. Certainly PageRank isn't much of a qualitative improvement over prior work like HITS, and any reasonable web-scale search engine can easily incorporate some variant on the HITS/PageRank idea.
Even better than that the 1965 Hubble paper is pretty relevant to the "folklore" algorithms for eigenvector based graph partitioning that keep getting re-discovered (without attribution) again and again.
PR is simple enough that "Given this link graph, what is the PR of site C" was a question on a written exam in my introductory AI class (meaning no notes, calculators, computers).
This reminds me of the minor hub-bub around PhotoSynth/VirtualEarth not being a new technology (e.g. academic researchers have previously researched it in the past, way before computers had any hope of executing the complex matrix transformations.)
