I wonder if the Decision Tree base estimator could derive from a more general 
base estimator for Random Forests and just, for example, override a setup 
method or a constructor?

-----Original Message-----
From: Andreas <[email protected]>
Date: Fri, 20 Jan 2012 22:53:40 
To: <[email protected]>
Reply-To: [email protected]
Subject: Re: [Scikit-learn-general] Ensemble meta-estimators

On 01/20/2012 10:45 PM, Gilles Louppe wrote:
> Yes indeed, as I said at the time, much of the forest code could be
> reused to implement a pure averaging meta-estimator.
>
> The main thing that makes BaseForest tree-specific is that it
> precomputes X_argsorted such that it is computed only once for all
> trees and inject it into the fit method of the base trees. But besides
> that, any base estimator could actually be used.
>
>    
That I overlooked. Thanks for pointing it out.

I was wondering whether we want the code there for
both, forests and meta estimators.
This seems a bit redundant.

The precomputation is certainly a major optimization.
This could be included in a meta estimator but it would
increase complexity quite a bit.
On the other hand, maybe there are other
estimators that can be boosted and would profit
from some prior computations. I don't know whether
this is actually the case, though.

What do you think?

Cheers,
Andy

------------------------------------------------------------------------------
Keep Your Developer Skills Current with LearnDevNow!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-d2d
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
------------------------------------------------------------------------------
Keep Your Developer Skills Current with LearnDevNow!
The most comprehensive online learning library for Microsoft developers
is just $99.99! Visual Studio, SharePoint, SQL - plus HTML5, CSS3, MVC3,
Metro Style Apps, more. Free future releases when you subscribe now!
http://p.sf.net/sfu/learndevnow-d2d
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to