DOC improved plot_semi_supervised_newsgroups.py example #31104

elhambbi · 2025-03-29T10:51:44Z

Reference Issues/PRs

Towards #30621 and with reference to PR #30882, the code for plot_semi_supervised_newsgroups.py is improved.

What does this implement/fix? Explain your changes.

The example and the methods used are described more
Old code is updated e.g. f-strings, prints etc
A bar plot is added to compare the F1 score of different approaches

Any other comments?

LabelSpreading, as a semi-supervised method, is not better than the fully supervised models in terms of F1 score. Should we keep it or is SelfTraining enough to demonstrate the superiority of semi-supervised methods when having limited data?

github-actions · 2025-03-29T10:52:53Z

❌ Linting issues

This PR is introducing linting issues. Here's a summary of the issues. Note that you can avoid having linting issues by enabling pre-commit hooks. Instructions to enable them can be found here.

You can see the details of the linting issues under the lint job here

`ruff check`

ruff detected issues. Please run ruff check --fix --output-format=full locally, fix the remaining issues, and push the changes. Here you can see the detected issues. Note that the installed ruff version is ruff=0.11.7.


examples/semi_supervised/plot_semi_supervised_newsgroups.py:182:11: W292 [*] No newline at end of file
    |
180 | plt.tight_layout()
181 | plt.subplots_adjust(bottom=0.15)
182 | plt.show()
    |           ^ W292
    |
    = help: Add trailing newline

Found 1 error.
[*] 1 fixable with the `--fix` option.

`ruff format`

ruff detected issues. Please run ruff format locally and push the changes. Here you can see the detected issues. Note that the installed ruff version is ruff=0.11.7.


--- examples/semi_supervised/plot_semi_supervised_newsgroups.py
+++ examples/semi_supervised/plot_semi_supervised_newsgroups.py
@@ -179,4 +179,4 @@
 
 plt.tight_layout()
 plt.subplots_adjust(bottom=0.15)
-plt.show()
\ No newline at end of file
+plt.show()

1 file would be reformatted, 922 files already formatted

_{Generated for commit: bd476af. Link to the linter CI: here}

elhambbi · 2025-03-31T18:19:26Z

Hi @StefanieSenger . Could you please review the modifications I've done to this example and let me know if we need to change something?
Thank you.

StefanieSenger

Thanks a lot for your work, @elhambbi. It's going to upgrade the example a lot!

I especially like the part in the beginning that explains the intend of the example and summarises the approaches that get compared to each other.

I have a few little comments and also it would be nice to make the code more notebook style like and introduce some sub-sections.

Would you mind doing that?

StefanieSenger · 2025-04-09T11:52:49Z

examples/semi_supervised/plot_semi_supervised_newsgroups.py

+1. Supervised learning using 100% of labeled data (baseline)
+   - Uses SGDClassifier with TF-IDF features
+   - Represents the best possible performance with full supervision


The bullet points don't get rendered as intended. Can you please fix this?
You can see the rendering by building the documentation locally or by clicking "check the rendered docs" in the CI.

examples/semi_supervised/plot_semi_supervised_newsgroups.py

StefanieSenger · 2025-04-09T12:08:05Z

examples/semi_supervised/plot_semi_supervised_newsgroups.py

+    ls_pipeline, X_train, y_train_semi, X_test, y_test
+)
+
+# Create the plot


I think that instead of having comments like # Create the plot, # Define colors for each bar, # Customize plot, or # Add value labels on bars, it would be nice to add 1-2 sentences on top of this part explaining what will happen in this next section which would adhere more to the notebook style that we thrive for with the examples. This is a good example to look at, if you need some inspiration, @elhambbi: #26956

StefanieSenger · 2025-05-06T11:49:37Z

Hi @elhambbi, did you have some time to look into this?

elhambbi · 2025-05-06T16:15:27Z

Hi @StefanieSenger. I apologize, I've been too busy lately. I'll work on it over the weekend

StefanieSenger · 2025-05-07T07:13:18Z

Hi @elhambbi, no rush. I just wanted to know whether this PR is still active. Take your time.

elhambbi · 2025-05-24T20:16:50Z

Hi @StefanieSenger. Sorry, I've been quite busy lately. I have modified the code. Please let me know if it needs further improvement. Thank you

StefanieSenger

Thank you @elhambbi!

I have only got one little suggestion on the notebook style and some nit picks. Overall it looks pretty good to me.

Feel free to do more of the notebook style-like changes, but from my side with having these two blocks I'm pretty happy and would then forward this to a maintainer to have a look.

StefanieSenger · 2025-05-27T08:26:04Z

examples/semi_supervised/plot_semi_supervised_newsgroups.py

+The example uses the 20 newsgroups dataset, focusing on five categories.
+The results demonstrate how semi-supervised methods can achieve better
+performance than supervised learning with limited labeled data by
+effectively utilizing unlabeled samples.



Suggested change

(Sorry, this is really a nit.) We don't need this line.

examples/semi_supervised/plot_semi_supervised_newsgroups.py

elhambbi · 2025-06-01T15:27:34Z

Hi @StefanieSenger . I made the changes, and the file is updated now. I closed the PR by mistake and then reopened it. The commit history is gone but the code is correct with all the changes we discussed. Thank you

ogrisel · 2025-06-02T15:17:03Z

@elhambbi thanks for the PR. Could you please push a fix to add the missing new line as reported by the linter?

https://app.circleci.com/pipelines/github/scikit-learn/scikit-learn/67920/workflows/b45c787a-c741-4361-9085-15c44eb03ecd/jobs/306979

This would help get the Continuous Integration to proceed with building the HTML preview of example edited by the PR to make it easier to review.

ogrisel · 2025-06-02T15:19:36Z

examples/semi_supervised/plot_semi_supervised_newsgroups.py


-You can adjust the number of categories by giving their names to the dataset
-loader or setting them to `None` to get all 20 of them.
+1. Supervised learning using 100% of labeled data (baseline)


In wouldn't call this a baseline: if we do not have access to 100% labeled data, it will probably not be possible to reach the performance of this model. I would rather call it a "best case scenario".

ogrisel · 2025-06-02T15:19:53Z

examples/semi_supervised/plot_semi_supervised_newsgroups.py

+   - Uses SGDClassifier with TF-IDF features
+   - Represents the best possible performance with full supervision
+
+2. Supervised learning using only 20% of labeled data


This on the other hand could be called a "baseline" to compare semi-supervised learning methods to.

ogrisel · 2025-06-02T15:27:05Z

examples/semi_supervised/plot_semi_supervised_newsgroups.py


-    # select a mask of 20% of the train dataset
-    y_mask = np.random.rand(len(y_train)) < 0.2
+# Evaluate supervised model with 100% of training data


We should probably split each call to eval_and_get_f1 in its own cell by using # %% delimiters. Then the comments such as # Evaluate supervised model with 100% of training data should be slightly rephrased as grammatically correct sentences or paragraphs instead of just short inline code comments as @StefanieSenger suggested above.

StefanieSenger · 2025-06-02T15:52:03Z

Thanks for your work, @elhambbi! I see this really taking shape and that a core developer is now reviewing this (not just me as a team member) shows the progress of this PR. (Please: for expectation management, what I wrote in the What comes next? section of #30621 applies here: a PR like that is expected to be thoroughly reviewed and the result will be something a new contributor can be very proud of once it is finished.)

I will review more later. Just a little comment on the commit history: it has disappeared because of force-pushing, which we try to avoid at all costs. PRs are easier to review when we can see the commit history and as someone coming back to this PR regularily, I can rely on only looking at the newest changes instead of going through everything from the beginning. (Because I only have so much mental capacity and I jump a lot between very different topics and PRs.)

github-actions bot added the Documentation label Mar 29, 2025

StefanieSenger reviewed Apr 9, 2025

View reviewed changes

StefanieSenger self-requested a review May 27, 2025 08:38

StefanieSenger reviewed May 27, 2025

View reviewed changes

elhambbi closed this Jun 1, 2025

elhambbi force-pushed the cont03 branch from 93bf70d to 4560abc Compare June 1, 2025 15:09

fixed style

bd476af

elhambbi reopened this Jun 1, 2025

ogrisel reviewed Jun 2, 2025

View reviewed changes

adrinjalali mentioned this pull request Jun 4, 2025

DOC add link to plot_semi_supervised_newsgroups.py example in semi_supervised.rst #30882

Closed

StefanieSenger mentioned this pull request Jun 4, 2025

Add links to examples from the docstrings and user guide #30621

Open

Uh oh!

DOC improved plot_semi_supervised_newsgroups.py example #31104

Are you sure you want to change the base?

DOC improved plot_semi_supervised_newsgroups.py example #31104

Conversation

elhambbi commented Mar 29, 2025

Reference Issues/PRs

What does this implement/fix? Explain your changes.

Any other comments?

Uh oh!

github-actions bot commented Mar 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

❌ Linting issues

ruff check

ruff format

Uh oh!

elhambbi commented Mar 31, 2025

Uh oh!

StefanieSenger left a comment • 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!

StefanieSenger Apr 9, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

StefanieSenger Apr 9, 2025

Choose a reason for hiding this comment

Uh oh!

StefanieSenger commented May 6, 2025

Uh oh!

elhambbi commented May 6, 2025

Uh oh!

StefanieSenger commented May 7, 2025

Uh oh!

elhambbi commented May 24, 2025

Uh oh!

StefanieSenger left a comment • 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!

StefanieSenger May 27, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

elhambbi commented Jun 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ogrisel commented Jun 2, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ogrisel Jun 2, 2025

Choose a reason for hiding this comment

Uh oh!

ogrisel Jun 2, 2025

Choose a reason for hiding this comment

Uh oh!

ogrisel Jun 2, 2025

Choose a reason for hiding this comment

Uh oh!

StefanieSenger commented Jun 2, 2025

Uh oh!

Uh oh!

github-actions bot commented Mar 29, 2025 •

edited

Loading

`ruff check`

`ruff format`

StefanieSenger left a comment •

edited

Loading

StefanieSenger left a comment •

edited

Loading

elhambbi commented Jun 1, 2025 •

edited

Loading

ogrisel commented Jun 2, 2025 •

edited

Loading