On Sun, Jun 2, 2013 at 6:08 PM, Mathieu Blondel <math...@mblondel.org>wrote:
>
>
> On Sun, Jun 2, 2013 at 4:26 PM, Joel Nothman <jnoth...@student.usyd.edu.au
> > wrote:
>
>>
>> That's only true if users know they are required to pass binarized input
>> to cross-validation routines such as GridSearchCV and cross_val_score, or
>> else they might land up with a 2d array of ints instead of a 1d array of
>> objects.
>>
>
> I hadn't thought of compatibility with grid search... We need to add tests
> for this in test_multiclass.py.
>
> Yes, or at least cross_val_score. I haven't taken a look in detail at
test_multiclass.py, but we at least should be testing for identical results
given: sequence of sequences, binarized with neg_label=0, binarized with
neg_label=-1. And we need to test it with a realistic dataset, and the edge
case where all entries have the same number of labels (and hence np.asarray
does something different). And if there are no fundamental failures in
this, perhaps sequences of sequences aren't so harmful.
- Joel
------------------------------------------------------------------------------
Get 100% visibility into Java/.NET code with AppDynamics Lite
It's a free troubleshooting tool designed for production
Get down to code-level detail for bottlenecks, with <2% overhead.
Download for free and get started troubleshooting in minutes.
http://p.sf.net/sfu/appdyn_d2d_ap2
_______________________________________________
Scikit-learn-general mailing list
Scikit-learn-general@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
