← → ***





PageRank

[You voted for me and who else?
Also, how popular are you?]

Bibliometrics - PageRank's influence

But... web page links != paper 'links'

PageRank - simple ('essence') version


https://patents.google.com/patent/US6285999B1/en

An example

An[other] example

You can play with PR here: https://bytes.usc.edu/~saty/tools/xem/run.html?x=PR

PageRank - fuller version [with 'MBI'!]

PR computation, via iteration

The last three slides might seem confusing - because they use the current (rather than previous) values of PR(A), to calculate PR(B). "Technically" we ARE supposed to be using PR(A) and PR(B) from the previous step, to update them for the current step. We can 'get away with' being careless/sloppy because... it doesn't matter (which we use - previous or current values)! After enough steps, either way will result in correct, converged PR values.

The following screengrabs illustrate the "proper" way of using prior values, but, with arbitrary starting values to show they don't matter.

Convergence

A bunch of examples [PR for different link graphs]

Suggs for getting a good PR

PR - just one of many 'signals' used by the Google SE

Learning more

Fun reading: