Discussion:
Fwd: y parameter in s.l.preprocessing
(too old to reply)
Paolo Losi
2011-05-09 09:41:37 UTC
Permalink
Sent from the "wrong" email address...


---------- Forwarded message ----------
From: Paolo Losi <***@enuan.com>
Date: Fri, May 6, 2011 at 7:07 PM
Subject: y parameter in s.l.preprocessing
To: scikit-learn-***@lists.sourceforge.net


Hi all,

When using preprocessing.Scaler as the first step
in a Pipeline for a classification problem,
I had to add y=None to the fit parameter in order to make it work.

I've seen that y=None has been removed in:

https://github.com/scikit-learn/scikit-learn/commit/4152157ddd36b2ce1a880508526ad5c369412bd0

What is the rational for removal?

Thanks!
Olivier Grisel
2011-05-09 09:50:13 UTC
Permalink
Post by Paolo Losi
Hi all,
When using preprocessing.Scaler as the first step
in a Pipeline for a classification problem,
I had to add y=None to the fit parameter in order to make it work.
https://github.com/scikit-learn/scikit-learn/commit/4152157ddd36b2ce1a880508526ad5c369412bd0
What is the rational for removal?
I think this is a mistake. We discovered that y=None was necessary to
use have gridsearch works at expected later on the PCA model for
instance. I would +1 a pull request to revert that change.
--
Olivier
http://twitter.com/ogrisel - http://github.com/ogrisel
Gael Varoquaux
2011-05-09 09:51:44 UTC
Permalink
Post by Olivier Grisel
Post by Paolo Losi
https://github.com/scikit-learn/scikit-learn/commit/4152157ddd36b2ce1a880508526ad5c369412bd0
What is the rational for removal?
I think this is a mistake. We discovered that y=None was necessary to
use have gridsearch works at expected later on the PCA model for
instance. I would +1 a pull request to revert that change.
+1.

G
Mathieu Blondel
2011-05-09 10:10:30 UTC
Permalink
On Mon, May 9, 2011 at 6:51 PM, Gael Varoquaux
Post by Olivier Grisel
Post by Paolo Losi
https://github.com/scikit-learn/scikit-learn/commit/4152157ddd36b2ce1a880508526ad5c369412bd0
What is the rational for removal?
I think this is a mistake. We discovered that y=None was necessary to
use have gridsearch works at expected later on the PCA model for
instance. I would +1 a pull request to revert that change.
+1

Mathieu
Olivier Grisel
2011-05-09 10:16:36 UTC
Permalink
Post by Mathieu Blondel
On Mon, May 9, 2011 at 6:51 PM, Gael Varoquaux
Post by Olivier Grisel
Post by Paolo Losi
https://github.com/scikit-learn/scikit-learn/commit/4152157ddd36b2ce1a880508526ad5c369412bd0
What is the rational for removal?
I think this is a mistake. We discovered that y=None was necessary to
use have gridsearch works at expected later on the PCA model for
instance. I would +1 a pull request to revert that change.
+1
BTW this is a great opportunity to add a new regression test to the
pipeline test case for such integration of the preprocessor with the
pipeline.
--
Olivier
http://twitter.com/ogrisel - http://github.com/ogrisel
Paolo Losi
2011-05-09 10:24:45 UTC
Permalink
On Mon, May 9, 2011 at 12:16 PM, Olivier Grisel
Post by Olivier Grisel
BTW this is a great opportunity to add a new regression test to the
pipeline test case for such integration of the preprocessor with the
pipeline.
Ok. I'll take care of it.
Mathieu Blondel
2011-05-09 11:07:12 UTC
Permalink
Post by Paolo Losi
Ok. I'll take care of it.
Merged! Thanks Paolo.

Mathieu

Loading...