You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@nutch.apache.org by Bartosz Gadzimski <ba...@o2.pl> on 2009/03/27 15:15:56 UTC

LinkRank why 10 iterations?

Hello,

Why you are making so many iterations in linkrank are this is neccessary 
for some amount of websites?

Thanks,
Bartosz

Re: LinkRank why 10 iterations?

Posted by Dennis Kubes <ku...@apache.org>.
You are running LinkRank and a comparatively small webgraph.  LinkRank 
is meant, in principle, to be run on very large webgraphs, millions or 
perhaps 100s of millions of urls.  On that scale 10 iterations was what 
we saw as a good default for the webgraph to converge while not taking 
an excessive amount of time.  For smaller webgraphs 10 iterations might 
not be necessary.

You can use the link.analyze.num.iterations configuration variable to 
set the number of iterations you would like to run.  As a general rule I 
don't think I would ever go below 5 iterations as all but the very 
smallest webgraphs wouldn't have enough chance to converge.

Here is a good paper on PageRank and convergence.  Principles are the same:

http://www.webworkshop.net/pagerank.html

Dennis



Bartosz Gadzimski wrote:
> Hello,
> 
> Why you are making so many iterations in linkrank are this is neccessary 
> for some amount of websites?
> 
> Thanks,
> Bartosz