Hi Andy,

On 27 December 2011 13:37, Andreas Müller <[email protected]> wrote:
> Hi everybody!
> I was wondering if anyone is working on ensemble meta-estimators.
> There seems to have been some effort by @glouppe when doing the random
> forests.
> Is that still going on?

I am not working on that currently.

> Also, I was wondering how tree-specific the random forest module is.
> I looked at the pull request but could not find much about this.
> Was there any discussion on this that I missed? What was the reasoning
> behind having mixins instread of meta-classifier for bagging?

The current implementation is not that much tree-specific, but I
implemented some optimizations in my pending pull requests #478 and
#491 that makes the BaseForest class highly tree-specific.

> Or do you think that bagging is so easy to implement that it
> does not pay off to have an extra meta estimator?

I think meta-estimators would a good thing to have and that much of
the forest code could be reused to implement a bagging meta-estimator.

Gilles

------------------------------------------------------------------------------
Write once. Port to many.
Get the SDK and tools to simplify cross-platform app development. Create 
new or port existing apps to sell to consumers worldwide. Explore the 
Intel AppUpSM program developer opportunity. appdeveloper.intel.com/join
http://p.sf.net/sfu/intel-appdev
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to