mahout-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Isabel Drost <>
Subject Re: GSOC
Date Sun, 30 Mar 2008 11:15:28 GMT
On Saturday 29 March 2008, Ted Dunning wrote:
> SVM is not the only solution to these problems.  For many search engine
> applications, it isn't even likely to be the best.  Regularized logistic
> regression is a strong candidate as are random forests and boosted trees.

There have been several interesting papers on ranking search results based on 
preferences on NIPS 2007. The algorithms presented therein optimise exactly 
the criterion used to evaluate search engine rankings. In some cases they 
also compare against the svm solution of Thorsten Joachims.

> The algorithm may well have some 
> virtues, but it is unlikely to be universal. 

There is even an "official theorem" for your statement: the no-free-lunch 
theorem :)

> It is more likely that the author who claims this simply has a limited view
> of the range of things that might need to be done.

Or that he just examined the algorithm from exactly one angle that might not 
be the one that is important for your problem.


Flattery is like cologne -- to be smelled, but not swallowed.		-- Josh 
  |\      _,,,---,,_       Web:   <>
  /,`.-'`'    -.  ;-;;,_
 |,4-  ) )-,_..;\ (  `'-'
'---''(_/--'  `-'\_) (fL)  IM:  <xmpp://>

View raw message