[MRG+2] ENH: (minor) add quantile_low/high parameters to robust scaler #5929

code-of-kpp · 2015-11-27T18:40:10Z

No description provided.

giorgiop · 2015-11-27T21:40:19Z

sklearn/preprocessing/data.py

@@ -911,6 +912,12 @@ class RobustScaler(BaseEstimator, TransformerMixin):
    with_scaling : boolean, True by default
        If True, scale the data to interquartile range.

+    quantile_low : float


Could you please add .. versionadded:: 0.18 where needed in the docstrings?

giorgiop · 2015-11-27T21:45:29Z

This looks like an useful addition to me. @GaelVaroquaux any opinion?
Also, whats_new might need to be updated.

code-of-kpp · 2015-11-30T10:16:31Z

Could you tests for extreme values of quantile_low/high? For example when quantile_low >= quantile_high and when quantile_low=0 etc. Also, does RobustScaler ever reduces to StandardScaler for any value of those parameters?

@giorgiop: Is it expected to raise at __init__ time, at set_params time or at fit time?

giorgiop · 2015-11-30T14:16:25Z

I think the best is at fit time. You could use MinMaxScaler as inspiration.

code-of-kpp · 2015-12-01T10:55:55Z

Also, does RobustScaler ever reduces to StandardScaler for any value of those parameters?

@giorgiop
In general it is not the case. But if you know original distribution of X and sample is large enough then you can calculate quantiles for mean +- std.

giorgiop · 2015-12-02T14:07:35Z

sklearn/preprocessing/data.py

@@ -911,6 +912,14 @@ class RobustScaler(BaseEstimator, TransformerMixin):
    with_scaling : boolean, True by default
        If True, scale the data to interquartile range.

+


cosmetics: only one blank line in the docstrings, same below.

giorgiop · 2015-12-02T14:50:02Z

For the rest, LGMT.

code-of-kpp · 2015-12-05T09:50:08Z

git blame shows some docstrings was added by @amueller, may be he will be interested in reviewing this.
hello

glouppe · 2016-02-11T09:56:10Z

sklearn/preprocessing/data.py

@@ -1,3 +1,4 @@
+


Please remove this blank line

glouppe · 2016-02-11T10:01:16Z

Seems like a fine addition to me. +1 for merge provided my comments are fixed.

(Sorry for the slow reply here...)

code-of-kpp · 2016-07-29T12:32:06Z

bump?

amueller · 2016-07-29T14:56:53Z

Sorry for the slow turnaround. LGTM.

amueller · 2016-07-29T14:57:13Z

Do you want to add an entry to whatsnew?

code-of-kpp · 2016-08-10T20:32:24Z

Sorry, didn't realize that was necessarily.
Do I have to rebase now or something?

nelson-liu · 2016-08-10T21:03:43Z

doc/whats_new.rst

@@ -59,6 +59,10 @@ Enhancements
    generated during build. Distribution packages will still contain generated
    C/C++ files. By `Arthur Mensch`_

+   - :class: `RobustScaler` now accepts ``quantile_range`` parameter.


remove the space between class: and RobustScaler to have it link properly

code-of-kpp · 2016-08-11T00:01:25Z

yep
and now we have conflicts

nelson-liu · 2016-08-11T00:04:38Z

Can you rebase on master?

code-of-kpp · 2016-08-11T00:19:58Z

I'm trying but strangely after rebase I find myself in detached HEAD

add versionadded tags

fix typo in docstrings of RobustScaler

add missing url link in whatsnew

code-of-kpp · 2016-08-11T00:43:09Z

Looks like it is ok now

agramfort · 2016-08-13T08:32:24Z

thanks @podshumok

scikit-learn#5929) * add quantile_low/high parameters to robust scaler add versionadded tags * robust scaler: add parameter for quantile_range, check its validity * robust scaler: simplify params validations fix typo in docstrings of RobustScaler * preprocessing.data fix some lines > 80 columns * add whatsnew for robust scaler quantile_range param add missing url link in whatsnew

giorgiop reviewed Nov 27, 2015
View reviewed changes

code-of-kpp changed the title ~~add quantile_low/high parameters to robust scaler~~ [NEEDS REVIEW] ENH: (minor) add quantile_low/high parameters to robust scaler Nov 30, 2015

giorgiop reviewed Dec 2, 2015
View reviewed changes

code-of-kpp changed the title ~~[NEEDS REVIEW] ENH: (minor) add quantile_low/high parameters to robust scaler~~ [MRG] ENH: (minor) add quantile_low/high parameters to robust scaler Dec 2, 2015

amueller added the Waiting for Reviewer label Dec 10, 2015

glouppe reviewed Feb 11, 2016
View reviewed changes

sklearn/preprocessing/data.py

@@ -1,3 +1,4 @@

Copy link

Contributor

glouppe Feb 11, 2016

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please remove this blank line

glouppe changed the title ~~[MRG] ENH: (minor) add quantile_low/high parameters to robust scaler~~ [MRG+1] ENH: (minor) add quantile_low/high parameters to robust scaler Feb 11, 2016

amueller changed the title ~~[MRG+1] ENH: (minor) add quantile_low/high parameters to robust scaler~~ [MRG+2] ENH: (minor) add quantile_low/high parameters to robust scaler Jul 29, 2016

nelson-liu reviewed Aug 10, 2016
View reviewed changes

code-of-kpp force-pushed the minor_enh_add_quantile_params_to_robust_scaler branch from 93991d5 to fe7b974 Compare August 11, 2016 00:00

code-of-kpp force-pushed the minor_enh_add_quantile_params_to_robust_scaler branch from fe7b974 to 54ea875 Compare August 11, 2016 00:31

Konstantin Podshumok and others added 5 commits August 11, 2016 03:36

add quantile_low/high parameters to robust scaler

2208eab

add versionadded tags

robust scaler: add parameter for quantile_range, check its validity

bcf1029

robust scaler: simplify params validations

1d38772

fix typo in docstrings of RobustScaler

preprocessing.data fix some lines > 80 columns

14f4e37

add whatsnew for robust scaler quantile_range param

fb252b0

add missing url link in whatsnew

code-of-kpp force-pushed the minor_enh_add_quantile_params_to_robust_scaler branch from a9137f4 to fb252b0 Compare August 11, 2016 00:41

agramfort merged commit f893565 into scikit-learn:master Aug 13, 2016

code-of-kpp deleted the minor_enh_add_quantile_params_to_robust_scaler branch August 13, 2016 17:33

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[MRG+2] ENH: (minor) add quantile_low/high parameters to robust scaler #5929

[MRG+2] ENH: (minor) add quantile_low/high parameters to robust scaler #5929

code-of-kpp commented Nov 27, 2015

giorgiop Nov 27, 2015

giorgiop commented Nov 27, 2015

code-of-kpp commented Nov 30, 2015

giorgiop commented Nov 30, 2015

code-of-kpp commented Dec 1, 2015

giorgiop Dec 2, 2015

giorgiop commented Dec 2, 2015

code-of-kpp commented Dec 5, 2015

glouppe Feb 11, 2016

glouppe commented Feb 11, 2016

code-of-kpp commented Jul 29, 2016

amueller commented Jul 29, 2016

amueller commented Jul 29, 2016

code-of-kpp commented Aug 10, 2016

nelson-liu Aug 10, 2016

code-of-kpp commented Aug 11, 2016

nelson-liu commented Aug 11, 2016

code-of-kpp commented Aug 11, 2016 •

edited

Loading

code-of-kpp commented Aug 11, 2016

agramfort commented Aug 13, 2016

		@@ -911,6 +912,14 @@ class RobustScaler(BaseEstimator, TransformerMixin):
		with_scaling : boolean, True by default
		If True, scale the data to interquartile range.

[MRG+2] ENH: (minor) add quantile_low/high parameters to robust scaler #5929

[MRG+2] ENH: (minor) add quantile_low/high parameters to robust scaler #5929

Conversation

code-of-kpp commented Nov 27, 2015

giorgiop Nov 27, 2015

Choose a reason for hiding this comment

giorgiop commented Nov 27, 2015

code-of-kpp commented Nov 30, 2015

giorgiop commented Nov 30, 2015

code-of-kpp commented Dec 1, 2015

giorgiop Dec 2, 2015

Choose a reason for hiding this comment

giorgiop commented Dec 2, 2015

code-of-kpp commented Dec 5, 2015

glouppe Feb 11, 2016

Choose a reason for hiding this comment

glouppe commented Feb 11, 2016

code-of-kpp commented Jul 29, 2016

amueller commented Jul 29, 2016

amueller commented Jul 29, 2016

code-of-kpp commented Aug 10, 2016

nelson-liu Aug 10, 2016

Choose a reason for hiding this comment

code-of-kpp commented Aug 11, 2016

nelson-liu commented Aug 11, 2016

code-of-kpp commented Aug 11, 2016 • edited Loading

code-of-kpp commented Aug 11, 2016

agramfort commented Aug 13, 2016

code-of-kpp commented Aug 11, 2016 •

edited

Loading