[MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() #8137

naoyak · 2016-12-30T04:50:39Z

Fixes #1975

Adds a DeprecationWarning if the ridge_alpha parameter is passed to SparsePCA.transform().

jnothman

Thanks

jnothman · 2016-12-31T14:07:04Z

sklearn/decomposition/sparse_pca.py

@@ -10,6 +10,8 @@
 from ..base import BaseEstimator, TransformerMixin
 from .dict_learning import dict_learning, dict_learning_online

+import warnings


standard library imports should come first

jnothman · 2016-12-31T14:19:15Z

sklearn/decomposition/sparse_pca.py

@@ -148,7 +150,8 @@ def transform(self, X, ridge_alpha=None):

        ridge_alpha: float, default: 0.01
            Amount of ridge shrinkage to apply in order to improve
-            conditioning.
+            conditioning. Note that as of 0.19 this parameter is deprecated and


can use .. deprecated sphinx markup

naoyak · 2016-12-31T23:49:11Z

Thanks @jnothman - made the changes

jnothman · 2016-12-31T23:47:47Z

sklearn/decomposition/sparse_pca.py

@@ -150,6 +152,10 @@ def transform(self, X, ridge_alpha=None):
            Amount of ridge shrinkage to apply in order to improve
            conditioning.

+            .. deprecated:: 0.18
+               This parameter will be removed in 0.20.
+               Specify `ridge_alpha` in the `SparsePCA` constructor.


single backticks don't do anything in RST. Use double backticks.

Gotcha, thanks.

jnothman · 2017-01-01T00:17:19Z

Really, you should set ridge_alpha to 'deprecated' by default, so that if a user explicitly passes ridge_alpha=None it will still raise the warnin

jnothman · 2017-01-01T00:18:05Z

Also, please check for a test where transform with ridge_alpha is used, and apply assert_warns in the test to silence the warning during tests but ensure it is raised

jnothman · 2017-01-01T00:18:24Z

Travis failures are not your fault.

naoyak · 2017-01-01T01:08:14Z

@jnothman should the deprecated param be ignored, or should it still work for now? I've seen both patterns elsewhere in the codebase.

jnothman · 2017-01-01T01:31:21Z

Where do you see the deprecated param ignored? Yes, it should still work for now.

…

On 1 January 2017 at 12:08, Naoya Kanai ***@***.***> wrote: @jnothman <https://github.com/jnothman> should the deprecated param be ignored, or should it still work for now? I've seen both patterns elsewhere in the codebase. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#8137 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEz63iLSOg4W46kKa5Ni7WRBJ9Cr8NQks5rNvyAgaJpZM4LYAA4> .

jnothman · 2017-01-01T01:31:32Z

But it shouldn't raise a warning unless it is used.

…

On 1 January 2017 at 12:31, Joel Nothman ***@***.***> wrote: Where do you see the deprecated param ignored? Yes, it should still work for now. On 1 January 2017 at 12:08, Naoya Kanai ***@***.***> wrote: > @jnothman <https://github.com/jnothman> should the deprecated param be > ignored, or should it still work for now? I've seen both patterns elsewhere > in the codebase. > > — > You are receiving this because you were mentioned. > Reply to this email directly, view it on GitHub > <#8137 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AAEz63iLSOg4W46kKa5Ni7WRBJ9Cr8NQks5rNvyAgaJpZM4LYAA4> > . >

naoyak · 2017-01-01T01:37:21Z

Where do you see the deprecated param ignored? Yes, it should still work
for now.

I saw it ignored in #7992: 2e79d88#r91984196 - but without being familiar with that issue I'm not sure if the context is different.

jnothman · 2017-01-01T02:11:08Z

The code you refer to wasn't accepted to be merged.

…

On 1 January 2017 at 12:37, Naoya Kanai ***@***.***> wrote: Where do you see the deprecated param ignored? Yes, it should still work for now. I saw it ignored in 2e79d88#r91984196 <2e79d88#r91984196> - but without being familiar with that issue I'm not sure if the context is different. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#8137 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEz66fL3llAObpzUXfzHlyS87wSlFelks5rNwNTgaJpZM4LYAA4> .

naoyak · 2017-01-01T02:19:16Z

Sorry, I think that comment was an outdated diff. But it looks like current master for that perplexity method still throws away the deprecated arg. https://github.com/scikit-learn/scikit-learn/blob/master/sklearn/decomposition/online_lda.py#L750

naoyak · 2017-01-01T03:53:24Z

sklearn/decomposition/tests/test_sparse_pca.py

@@ -70,6 +71,10 @@ def test_fit_transform():
    spca_lasso.fit(Y)
    assert_array_almost_equal(spca_lasso.components_, spca_lars.components_)

+    # Test calling tranform with deprecated ridge_alpha parameter


Seeing the note about testing times with SparsePCA I stuck the deprecation test in here. Lmk if you think it's better to break it out.

*transform

Here is fine. It's frustrating that this parameter was not already tested.

jnothman · 2017-01-01T09:40:47Z

Sorry, I think that comment was an outdated diff. But it looks like current master for that perplexity method still throws away the deprecated arg. https://github.com/scikit-learn/scikit-learn/blob/ master/sklearn/decomposition/online_lda.py#L750

You're quite right. That needs to be fixed. A PR is welcome. That's also good motivation for making sure to test deprecate depredations. The problem there might have been that that parameter was never tested! :\

jnothman

Otherwise LGTM

jnothman · 2017-01-01T09:47:35Z

sklearn/decomposition/tests/test_sparse_pca.py

@@ -70,6 +71,10 @@ def test_fit_transform():
    spca_lasso.fit(Y)
    assert_array_almost_equal(spca_lasso.components_, spca_lars.components_)

+    # Test calling tranform with deprecated ridge_alpha parameter


*transform

Here is fine. It's frustrating that this parameter was not already tested.

jnothman · 2017-01-01T09:48:10Z

sklearn/decomposition/tests/test_sparse_pca.py

@@ -70,6 +71,10 @@ def test_fit_transform():
    spca_lasso.fit(Y)
    assert_array_almost_equal(spca_lasso.components_, spca_lars.components_)

+    # Test calling tranform with deprecated ridge_alpha parameter
+    ridge_alpha = 0.01
+    assert_warns(DeprecationWarning, spca_lars.transform, Y, ridge_alpha)


I'd rather ridge_alpha=0.01 there, in part because really transform should accept a second positional arg.

Also added an example where ridge_alpha=None to make sure it's handled properly

jnothman

but then I found more...

jnothman · 2017-01-01T09:49:38Z

sklearn/decomposition/sparse_pca.py

+                          "deprecated since 0.19 and will be removed in 0.21. "
+                          "Specify ridge_alpha in the SparsePCA constructor.",
+                          DeprecationWarning)
+        else:


This is going to break if ridge_alpha=None is passed.

Please test this case.

Thanks for the catch. In this case I have it reverting to self.ridge_alpha.

naoyak · 2017-01-01T11:12:34Z

@jnothman thanks for the detailed check, should be good now. Tests working again!

jnothman · 2017-01-01T11:26:51Z

Yes, again LGTM

tguillemot

+1 as my comment is not really important.

tguillemot · 2017-01-03T10:57:20Z

sklearn/decomposition/tests/test_sparse_pca.py

@@ -70,6 +71,10 @@ def test_fit_transform():
    spca_lasso.fit(Y)
    assert_array_almost_equal(spca_lasso.components_, spca_lars.components_)

+    # Test that deprecated ridge_alpha parameter throws warning
+    assert_warns(DeprecationWarning, spca_lars.transform, Y, ridge_alpha=0.01)
+    assert_warns(DeprecationWarning, spca_lars.transform, Y, ridge_alpha=None)


Just a niptick, it will be better to use assert_warns_message with a part of the message (just to be sure it's the right message).

Thanks for the tip

jnothman · 2017-01-04T08:42:13Z

Thanks @naoyak!

tguillemot · 2017-01-04T08:54:17Z

Thx @naoyak !!!

…learn#8137)

Deprecate ridge_alpha param on SparsePCA.transform()

fe539bc

jnothman reviewed Dec 31, 2016

View reviewed changes

Use deprecated sphinx syntax for SparsePCA docs

732c46c

naoyak force-pushed the sparse-pca-transform branch from b296690 to 732c46c Compare December 31, 2016 23:48

jnothman reviewed Dec 31, 2016

View reviewed changes

naoyak force-pushed the sparse-pca-transform branch from 7f08889 to 409509b Compare January 1, 2017 03:50

naoyak commented Jan 1, 2017

View reviewed changes

naoyak changed the title ~~Deprecate ridge_alpha param on SparsePCA.transform()~~ [MRG] Deprecate ridge_alpha param on SparsePCA.transform() Jan 1, 2017

Added test for SparsePCA transform param deprecation

806901a

naoyak force-pushed the sparse-pca-transform branch from 409509b to 806901a Compare January 1, 2017 05:56

jnothman reviewed Jan 1, 2017

View reviewed changes

jnothman changed the title ~~[MRG] Deprecate ridge_alpha param on SparsePCA.transform()~~ [MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() Jan 1, 2017

jnothman requested changes Jan 1, 2017

View reviewed changes

Check that ridge_alpha is not None on SparsePCA.transform()

442c404

naoyak mentioned this pull request Jan 2, 2017

[MRG + 1] Fix perplexity method by adding _unnormalized_transform method, Issue #7954 #7992

Merged

tguillemot approved these changes Jan 3, 2017

View reviewed changes

Check deprecation message in SparsePCA test

7cfe756

naoyak force-pushed the sparse-pca-transform branch from c60db80 to 7cfe756 Compare January 3, 2017 20:26

jnothman approved these changes Jan 4, 2017

View reviewed changes

jnothman merged commit 84349a7 into scikit-learn:master Jan 4, 2017

naoyak deleted the sparse-pca-transform branch January 4, 2017 08:45

raghavrv pushed a commit to raghavrv/scikit-learn that referenced this pull request Jan 5, 2017

[MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() (scikit-…

81f5b3e

…learn#8137)

sergeyf pushed a commit to sergeyf/scikit-learn that referenced this pull request Feb 28, 2017

[MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() (scikit-…

2edc335

…learn#8137)

Przemo10 mentioned this pull request Mar 17, 2017

update fork (#1) #8606

Closed

Sundrique pushed a commit to Sundrique/scikit-learn that referenced this pull request Jun 14, 2017

[MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() (scikit-…

8312a92

…learn#8137)

NelleV pushed a commit to NelleV/scikit-learn that referenced this pull request Aug 11, 2017

[MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() (scikit-…

674c8d9

…learn#8137)

paulha pushed a commit to paulha/scikit-learn that referenced this pull request Aug 19, 2017

[MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() (scikit-…

3458f2e

…learn#8137)

maskani-moh pushed a commit to maskani-moh/scikit-learn that referenced this pull request Nov 15, 2017

[MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() (scikit-…

fa8e1eb

…learn#8137)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() #8137

[MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() #8137

naoyak commented Dec 30, 2016 •

edited

Loading

jnothman left a comment

jnothman Dec 31, 2016

jnothman Dec 31, 2016

naoyak commented Dec 31, 2016

jnothman Dec 31, 2016

naoyak Jan 1, 2017

jnothman commented Jan 1, 2017

jnothman commented Jan 1, 2017

jnothman commented Jan 1, 2017

naoyak commented Jan 1, 2017

jnothman commented Jan 1, 2017 via email

jnothman commented Jan 1, 2017 via email

naoyak commented Jan 1, 2017 •

edited

Loading

jnothman commented Jan 1, 2017 via email

naoyak commented Jan 1, 2017

naoyak Jan 1, 2017

jnothman Jan 1, 2017

jnothman commented Jan 1, 2017 via email •

edited

Loading

jnothman left a comment

jnothman Jan 1, 2017

jnothman Jan 1, 2017

naoyak Jan 1, 2017

jnothman left a comment

jnothman Jan 1, 2017

jnothman Jan 1, 2017

naoyak Jan 1, 2017

naoyak commented Jan 1, 2017

jnothman commented Jan 1, 2017

tguillemot left a comment

tguillemot Jan 3, 2017

naoyak Jan 3, 2017

jnothman commented Jan 4, 2017

tguillemot commented Jan 4, 2017

[MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() #8137

[MRG+1] Deprecate ridge_alpha param on SparsePCA.transform() #8137

Conversation

naoyak commented Dec 30, 2016 • edited Loading

jnothman left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

naoyak commented Dec 31, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jnothman commented Jan 1, 2017

jnothman commented Jan 1, 2017

jnothman commented Jan 1, 2017

naoyak commented Jan 1, 2017

jnothman commented Jan 1, 2017 via email

jnothman commented Jan 1, 2017 via email

naoyak commented Jan 1, 2017 • edited Loading

jnothman commented Jan 1, 2017 via email

naoyak commented Jan 1, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jnothman commented Jan 1, 2017 via email • edited Loading

jnothman left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jnothman left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

naoyak commented Jan 1, 2017

jnothman commented Jan 1, 2017

tguillemot left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jnothman commented Jan 4, 2017

tguillemot commented Jan 4, 2017

naoyak commented Dec 30, 2016 •

edited

Loading

naoyak commented Jan 1, 2017 •

edited

Loading

jnothman commented Jan 1, 2017 via email •

edited

Loading