On Tue, Apr 17, 2012 at 10:48 PM, Andreas Mueller
<[email protected]>wrote:
> **
> Am 17.04.2012 15:45, schrieb Mathieu Blondel:
>
>
>
> On Tue, Apr 17, 2012 at 10:31 PM, Olivier Grisel <[email protected]
> > wrote:
>
>>
>> 1- use C and scale_C=False by default and document extensively the
>> importance of scale_C=True when doing model selection with small
>> number of samples. (I am ok for the ugly warning in the grid search
>> class).
>>
>
> Setting scale_C to None by default and raising a warning in fit whenever
> scale_C is still None is a good way to educate users. If they want the
> warning to go away, they'll have to make an informed choice and explicitly
> set scale_C to either False or True. We had this discussion several times
> and couldn't reach a consensus so letting users choose for themselves seems
> like the best bet (the warning message should help users make the best
> decision depending on their need).
>
>
> At least Dimitros did not seem to notice the warning about the behaviour
> of scale_C he got. Or maybe it was not informative?
>
>
The master branch raises a warning when scale_C=False, not when scale_C is
None... For me, that doesn't make sense at all: people who explicitly set
scale_C to False do it on purpose, so no need to raise a warning...
Mathieu
------------------------------------------------------------------------------
Better than sec? Nothing is better than sec when it comes to
monitoring Big Data applications. Try Boundary one-second
resolution app monitoring today. Free.
http://p.sf.net/sfu/Boundary-dev2dev
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
