I'm +1 for adding tests to ensure grid search meets usages that fall
outside of the strict domains of scikit-learn's estimators. If users that
apply it to problems of other shape (additional args, etc.) can write
tests, or state their requirements, I think that would be valuable in
ensuring GridSea
A sklearn_theano example ran into this problem while using cross_val_score.
At the time, `cross_val_score` was calling `check_arrays` - now it is
calling `indexable` and everything seems to go through fine again.
Michael
On Thu, Dec 18, 2014 at 10:59 PM, Robert McGibbon
wrote:
>
> +1 on allowing
+1 on allowing non-2d data arrays in GridSearchCV. We use this capability
in some of our model selection problems as well, to layer GridSearchCV on
top of our modified estimators that are designed to take other shapes.
On Thursday, December 18, 2014, Gael Varoquaux <
gael.varoqu...@normalesup.org>
On Thu, Dec 18, 2014 at 04:41:25PM -0500, Andy wrote:
> I'm pretty sure currently there is less restrictions, which makes sense
> for the gridsearch and cross-validation.
We are on the same page. These are fairly general utilities.
G
-
Lars,
I will discuss a feature_shape parameter with my collaborators.
Alex,
An "allow_nd" parameter for these functions would also be a great feature.
On Thu, Dec 18, 2014 at 4:18 PM, Gael Varoquaux <
gael.varoqu...@normalesup.org> wrote:
>
> On Thu, Dec 18, 2014 at 10:15:18PM +0100, Lars Buiti
On 12/18/2014 02:24 PM, Lars Buitinck wrote:
> 2014-12-18 20:14 GMT+01:00 David Brough :
>> Thank you for the quick response.
>> I currently using version 0.15.2. Our arrays have dimensions [n_samples, x,
>> y, z]. Below are the two trace-backs I get for both train_test_split and
>> GrindSearchCV.
On Thu, Dec 18, 2014 at 10:15:18PM +0100, Lars Buitinck wrote:
> > there is an "allow_nd" param I added to support this use case. Maybe
> > we could add it to GridSearchCV...
> We decided against this in the case of k-NN, where the suggestion (I
> don't remember by whom) was to alllow >2d for cust
2014-12-18 21:45 GMT+01:00 Alexandre Gramfort
:
> there is an "allow_nd" param I added to support this use case. Maybe
> we could add it to GridSearchCV...
We decided against this in the case of k-NN, where the suggestion (I
don't remember by whom) was to alllow >2d for custom distance metrics.
I'
there is an "allow_nd" param I added to support this use case. Maybe
we could add it to GridSearchCV...
Alex
--
Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server
from Actuate! Instantly Supercharge Your Bu
2014-12-18 20:14 GMT+01:00 David Brough :
> Thank you for the quick response.
> I currently using version 0.15.2. Our arrays have dimensions [n_samples, x,
> y, z]. Below are the two trace-backs I get for both train_test_split and
> GrindSearchCV. They both end at the same line of code.
The dimens
Hi Andy,
Thank you for the quick response.
I currently using version 0.15.2. Our arrays have dimensions [n_samples, x,
y, z]. Below are the two trace-backs I get for both train_test_split and
GrindSearchCV. They both end at the same line of code.
--
Hi David.
Sorry about the issue you are seeing.
Which version of scikit-learn are you using, and what are the input
types and shapes?
I don't think that there are any requirements on the dimensions in
current master.
There was a refactoring of input validation that made that a bit tricky
but
Hi,
I am working on developing a python package that uses machine learning
speed up the optimization of materials development (pymks.org). This
package is built on top of sklearn. We have an example in our documentation
where we have used train_test_split and GridSearchCV to search the
parameter s
13 matches
Mail list logo
