Hi,
On 17 November 2013 22:45, Thomas Dent <[email protected]> wrote:
> Hi Gilles -
>
> thanks for the reply. I think changing the relative class weights does
> more or less what we want, which is to optimize the classification at very
> low false alarm probability.
>
> Another question on the DecisionTreeClassifier, does the argument
>
> splitter='best'
>
> actually do anything? The RandomForest objects don't have it …
>
This parameter allows for switching between the random forest and the
extra-trees algorithm. This is why RandomForestClassifier does not have it.
It is enforced by the estimator.
Best,
Gilles
>
> Best, Thomas
>
> > Hi Thomas,
> >
> > Indeed, gini and entropy are the only supported impurity criteria for
> > classification. I don't think we have plans right now to add others -
> which
> > one do you have in mind?
> >
> > > how feasible would it be to have the option of passing custom function
> to
> > the tree or forest to use in splitting?
> >
> > This is not that easy, but impurity criteria are fairly well decoupled
> from
> > the tree code. You could experiment with other criteria provided that you
> > implement the (cython) Criterion interface as defined at
> >
> >
> https://github.com/glouppe/scikit-learn/blob/master/sklearn/tree/_tree.pyx#L48
> >
> >
> > > I see that sample weighting has recently been implemented for trees: I
> > would like to know how far the behaviour of the splitting can be
> influenced
> > by assigning the training data different weights for different classes
> > (which might effectively lead to an asymmetric function)
> >
> > Well, basically, this amounts to compute the node impurity as if you had
> > several copies of a same training sample. If a training sample has more
> > weight, then it will account for more in the calculation and vice-versa.
> In
> > particular, an interesting use of sample weights is when your
> > classification problem is unbalanced and that you make it virtually
> > balanced.
> >
> > Hope this helps,
> >
> > Gilles
> >
>
> --
> -----------------------------------------
> Institute for Gravitational Physics
> (Albert Einstein Institute)
> Callinstr. 38
> D-30167 Hannover, Germany
>
>
>
> ------------------------------------------------------------------------------
> DreamFactory - Open Source REST & JSON Services for HTML5 & Native Apps
> OAuth, Users, Roles, SQL, NoSQL, BLOB Storage and External API Access
> Free app hosting. Or install the open source package on any LAMP server.
> Sign up and see examples for AngularJS, jQuery, Sencha Touch and Native!
> http://pubads.g.doubleclick.net/gampad/clk?id=63469471&iu=/4140/ostg.clktrk
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>
------------------------------------------------------------------------------
DreamFactory - Open Source REST & JSON Services for HTML5 & Native Apps
OAuth, Users, Roles, SQL, NoSQL, BLOB Storage and External API Access
Free app hosting. Or install the open source package on any LAMP server.
Sign up and see examples for AngularJS, jQuery, Sencha Touch and Native!
http://pubads.g.doubleclick.net/gampad/clk?id=63469471&iu=/4140/ostg.clktrk
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
