hi,

a suggestion is to make sure the "simplest" models appear first on the
grid so it's the simplest model that gets selected.

Alex

On Wed, Mar 14, 2012 at 11:04 AM, Emanuele Olivetti
<[email protected]> wrote:
> Hi, As far as I understand GridSearchCV selects the parameter value with 
> lower score but
> in case of multiple parameter values having the same minimum score - which is 
> not
> infrequent in the case of small datasets - it always selects 
> deterministically the first
> one of the group. This is usually the one with lowest value among equivalents 
> because we
> usually feed ordered parameter grids to GridSearchCV. Isn't this a biased way 
> of doing the
> selection? What about picking one at random (among equivalents) each time in 
> order avoid
> bias through stochasticity? The related code is here:
> https://github.com/scikit-learn/scikit-learn/blob/master/sklearn/grid_search.py#L352
>  Best,
> Emanuele
>
>
> ------------------------------------------------------------------------------
> Virtualization & Cloud Management Using Capacity Planning
> Cloud computing makes use of virtualization - but cloud computing
> also focuses on allowing computing to be delivered as a service.
> http://www.accelacomm.com/jaw/sfnl/114/51521223/
> _______________________________________________
> Scikit-learn-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing 
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general

Reply via email to