PageRank Algorithm Applied To the Food Web

An anonymous reader brings word of a new application for PageRank, Google's link analysis algorithm: monitoring the food web in an ecosystem. A team of researchers found that a modified version of PageRank can predict with great accuracy which species are vital to the existence of others. Quoting: "Every species is embedded in a complex network of relationships with others. A single extinction can cascade into the loss of seemingly unrelated species. Investigating when this might happen using more conventional methods is complicated, as even in simple ecosystems, the number of combinations exceeds the number of atoms in the universe. So, it would be impossible to try them all. Co-author Dr. Stefano Allesina realized he could apply PageRank to the problem when he stumbled across an article in a journal of applied mathematics describing the Google algorithm. 'First of all, we had to reverse the definition of the algorithm. In PageRank, a web page is important if important pages point to it. In our approach, a species is important if it points to important species.'"

Re:But solve the real problem?

[macraig]

If you're under 40 and in good health you're in for a rather rude awakening, with those beliefs. You'll likely live to see the shit begin to hit the fan in a serious way. Google and other online sources should be education enough for you. I'm too disorganized to do anything more than plant the seed; you'll have to water and feed it.

Re:Very helpful for Endangered Species Act

[icegreentea]

People generally don't want to spend their money saving some worm. As it's been suggested before, the focus on megafauna partly exists just because they are charismatic. Panda bears are fricking cute. And people will donate money to save panda bears, their habitat, and everything else living there. It's unfortunate, but it's just facing the reality of the situation. Being able to identity the real keystone species would be great for the actual scientists, conservationists, and policy makers. But when its time to get funding, donations and support especially from the public, you can bet it'll be some megafauna. I mean, look at that cute widdy tiger cub. Worm eggs? UGH.

Eigenvector Centrality Measure

[sugarmotor]

The pagerank algorithm is better understood as a kind of Eigenvector Centrality Measure.

See http://en.wikipedia.org/wiki/Eigenvector_centrality#Eigenvector_centrality [wikipedia.org]

Meaning it is not new as a method by itself, but applying it to the linking-structure of the WWW in order to produce
relevant documents for a query, was new. I think it is fair to say that the Google Pagerank matters very little, outside of being able to rank otherwise
  not-comparable search results.

And so it is better to state that "a specialized Eigenvector Centrality Measure can predict with great accuracy which species are vital to the existence of others" instead of "a modified version of PageRank can predict with great accuracy which species are vital to the existence of others". One can see that also when one realizes that these biologists have no query, no search, no equivalent of search keywords.

On the other hand, when the post says "Co-author Dr. Stefano Allesina realized he could apply PageRank to the problem when he stumbled across an article in a journal of applied mathematics describing the Google algorithm." -- I guess he might have found the method through the Google name.

Stephan

Re:Why is this surprising?

[jpmorgan]

Yes, it is notable that this wasn't published in 'some mathematics journal'. Pagerank is computing the limiting distribution of a discrete-time markov chain applied to webpages using a certain statistical model of hyperlinks. It makes no sense to talk of applying 'pagerank' to things other than webpages, because that's what makes pagerank special! As soon as you take pagerank out of the web-context, it's just a steady state analysis of a markov chain, which is a standard statistical technique covered in undergraduate statistics courses. It's like saying applying bayesian inference to a problem in ecology is using a 'spam filter.'

For me, this tells me that perhaps these researchers should wander over to their local mathematics department more often. They might find all sorts of goodies that mathematicians have developed in the past few centuries. Dr. Allesina might have discovered that there was no need to reverse engineer the algorithm, since the underlying mathematical principles have been well understood for over a hundred years. We might have a better understanding of the world if most sciences took mathematical models as seriously as physicists do.

Re:Importance

[Onymous Coward]

I was worried about this. Either underspecifying "importance" or using it in a simplistic way (though you may not actually be doing these things).

It measures a kind of importance. Not importance in all respects. Specifically, it measures importance in interdependence. Which only very roughly translates to an idealized general or universal importance.

Remove humans from the web and you won't get much "impact on [the] ecosystem" (in the form of cascading die off). Yet humans are generally regarded as "important".

It should be noted (now and in each subsequent discussion of this topic) that the value of a species needs to be assessed by other measures as well. Like how beautiful its plumage is.