DOC use notebook-style for plot_faces_decomposition #22452

Valavanca · 2022-02-11T14:36:15Z

Reference Issues/PRs

Fix notebook-style examples #22406
- examples/decomposition/plot_faces_decomposition.py

What does this implement/fix? Explain your changes.

fix notebook-style and add more descriptions
small refactoring for the plot function

Any other comments?

HTML in the doc

glemaitre · 2022-02-11T15:12:41Z

examples/decomposition/plot_faces_decomposition.py


 for name, estimator, center in estimators:
-    print("Extracting the top %d %s..." % (n_components, name))
+    print("- Extracting the top %d '%s'..." % (n_components, name))


I propose to break this for loop and instead create a cell where:

we define the model

and plot the gallery

It will have the benefit to remove the if/else for switching the behaviour depending of the estimator.

I like the idea. Done in 43d4a5b.

examples/decomposition/plot_faces_decomposition.py

glemaitre · 2022-02-11T15:13:33Z

examples/decomposition/plot_faces_decomposition.py

-# #############################################################################
-# Various positivity constraints applied to dictionary learning.
-estimators = [
+dict_estimators = [


Could you as well break the for loop into three different cells?

Also done in 43d4a5b.

examples/decomposition/plot_faces_decomposition.py

Valavanca · 2022-02-13T12:51:50Z

Due to the changed order of generated images, it was also necessary to change the links to the images in the decomposition documentation.

glemaitre · 2022-02-15T09:36:47Z

examples/decomposition/plot_faces_decomposition.py

+# subcomponents that are maximally independent.
+
+# %%
+ica_estimator = decomposition.FastICA(n_components=n_components, whiten=True)


We can remove the FutureWarning.

Suggested change

ica_estimator = decomposition.FastICA(n_components=n_components, whiten=True)

ica_estimator = decomposition.FastICA(

n_components=n_components, whiten='arbitrary-variance'

)

Regarding the ConvergenceWarning, could you increase the number of iterations (or maybe increase the tolerance) such that we don't get this warning anymore.

Fixed. increased max_iter and tol

glemaitre · 2022-02-15T09:37:22Z

examples/decomposition/plot_faces_decomposition.py

+# Sparse comp. - MiniBatchSparsePCA
+# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


Suggested change

# Sparse comp. - MiniBatchSparsePCA

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

# Sparse components - MiniBatchSparsePCA

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

glemaitre · 2022-02-15T09:37:29Z

examples/decomposition/plot_faces_decomposition.py

+
+# %%
+# Independent components - FastICA
+# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


Suggested change

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Fixed all similar problems

glemaitre · 2022-02-15T09:37:36Z

examples/decomposition/plot_faces_decomposition.py

+
+# %%
+# Non-negative components - NMF
+# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


Suggested change

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

glemaitre · 2022-02-15T09:37:41Z

examples/decomposition/plot_faces_decomposition.py

+
+# %%
+# Eigenfaces - PCA using randomized SVD
+# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


Suggested change

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

glemaitre · 2022-02-15T09:39:11Z

examples/decomposition/plot_faces_decomposition.py

+# Dictionary learning - positive dictionary
+# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


Suggested change

# Dictionary learning - positive dictionary

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

# Dictionary learning - positive dictionary

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

glemaitre · 2022-02-15T09:39:18Z

examples/decomposition/plot_faces_decomposition.py

+# Dictionary learning - positive code
+# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


Suggested change

# Dictionary learning - positive code

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

# Dictionary learning - positive code

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

glemaitre · 2022-02-15T09:39:26Z

examples/decomposition/plot_faces_decomposition.py

+# Dictionary learning - positive dictionary & code
+# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


Suggested change

# Dictionary learning - positive dictionary & code

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

# Dictionary learning - positive dictionary & code

# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

glemaitre · 2022-02-15T09:43:18Z

examples/decomposition/plot_faces_decomposition.py

+# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+#
+
+# %%


In this section and the subsequent ones, it could be cool to add a small summary of what constraint we imposed in the model and how it translates into the components. No need to add long discussion but this is currently slightly rough.

I have added a bit of a description to the dictionary learning cells. But there is a slight nuance that I didn't find a comprehensive explanation of code in the documentation. But as far as I understand it is coding coefficients from sparse coding. Added so for now.

glemaitre · 2022-02-15T09:45:53Z

examples/decomposition/plot_faces_decomposition.py

+    n_components=n_components, svd_solver="randomized", whiten=True
+)
+pca_estimator.fit(faces_centered)
+plot_gallery(


I think that we should probably plot a colorbar with the min/max values.
Without it, I find it complex to grasp that a lot of components are zeros in SparsePCA for instance. I could have the same comment with only positive components.

So the colorbar could solve those ambuiguities.

Added colorbar, now plots look more informative.

lesteve · 2022-02-15T16:25:28Z

Due to the changed order of generated images, it was also necessary to change the links to the images in the decomposition documentation.

Oh wow, good catch @Valavanca! This is quite easy to overlook and definitely something to keep in mind for other PRs on this issue.

glemaitre · 2022-03-02T13:00:45Z

Thanks @Valavanca The example is much nicer now.

fix plots, use notebook style

72f76b7

github-actions bot added the Documentation label Feb 11, 2022

glemaitre reviewed Feb 11, 2022

View reviewed changes

Valavanca added 2 commits February 13, 2022 12:57

break loops into separate cells

43d4a5b

fix img links in doc

0138724

Valavanca requested a review from glemaitre February 14, 2022 13:11

glemaitre reviewed Feb 15, 2022

View reviewed changes

lesteve mentioned this pull request Feb 15, 2022

Fix notebook-style examples #22406

Closed

47 tasks

Valavanca added 2 commits February 20, 2022 18:12

fix markdown, add colorbar and desc for dict learning

86ad843

fix title

d0cb84b

lesteve added the Quick Review For PRs that are quick to review label Feb 22, 2022

Valavanca added 2 commits February 22, 2022 14:10

fix math symbol

8f70687

fix doc link

20ea3c5

glemaitre merged commit 2577dc9 into scikit-learn:main Mar 2, 2022

		# Sparse comp. - MiniBatchSparsePCA
		# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

	# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
	# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

	# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
	# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

	# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
	# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

		# Dictionary learning - positive dictionary
		# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

		# Dictionary learning - positive code
		# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Uh oh!

DOC use notebook-style for plot_faces_decomposition #22452

DOC use notebook-style for plot_faces_decomposition #22452

Uh oh!

Conversation

Valavanca commented Feb 11, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Reference Issues/PRs

What does this implement/fix? Explain your changes.

Any other comments?

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Valavanca Feb 13, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Valavanca commented Feb 13, 2022

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Valavanca Feb 20, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lesteve commented Feb 15, 2022

Uh oh!

glemaitre commented Mar 2, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Valavanca commented Feb 11, 2022 •

edited

Loading

Valavanca Feb 13, 2022 •

edited

Loading

Valavanca Feb 20, 2022 •

edited

Loading