As a part of #26927 we're also removing some examples or improving them when needed.

ping @ArturoAmorQ who is interested in this topic and might already have looked at aggregating some info.

+1 on removing examples then. We already have a way to redirect HTML pages as well.

@glemaitre as I mentioned also in #27108 (comment) , I'm not sure how much we should care about the redirections. I'd be totally okay with giving 404, especially since the only way users could get that 404 would be if they changed the URL manually. We don't provide any way for users to see the same page in a different version, and caring about a good redirect is making it really hard to remove examples.

Hi @lorentzenchr,

I would like to work on clustering examples, having reviewed, my observations and proposed changes are as follows:

• Agglomerative clustering with and without structure and Agglomerative clustering with different metrics​ are essentially similar examples, would like to keep the metrics example as it would seem more useful and covers enough context on working with agglomreative clustering.
• The example Feature agglomeration vs. univariate selection covers additional content on feature selection than the Feature agglomeration example, so we could remove the latter.
• K-means Clustering - There are already multiple comparative examples that show the difference between regular k-means and other variations like bisecting k-means and mini-batch k-means, so this example could be removed.

Could you advise if these would suffice or there is room to remove more.

(#29937)

These sound good suggestions @saldanhad

Hi @adrinjalali, appreciate your feedback, will submit a PR for these suggested changes soon.

Here is a non exhaustive list of other examples to be addressed:

• The Digit Dataset example can be merged with the Digit Classification example or even suppressed. DOC remove example showing a single image from digits dataset #29931
• Train error vs Test error example is really similar to Plotting Validation Curves example.
• SVM Margins example is really similar to SVM Maximal margin example and Support vectors in Linear SVC. Either of them could be merged with SVM separating hyperplane for umbalanced classes.
• The Feature importances with a forest of trees example demonstrates MDI for Random Forests. This is also covered in the Permutation importance vs MDI example and in the Gradient Boosting Regression example.
• Comparison of Calibration of Classifiers and Probability Calibration curves could be merged.
• Lasso and Elastic Net can be merged with Lasso path using Lars example. DOC Merge Lasso Lasso-LARS and ElasticNet examples together #29955
• Plot randomly generated classification dataset is nice for a visual intuition on the generated datasets, but could as well be replaced by snippets in the respective docstrings. DOC: Remove repeated/obselete examples under clustering. #29937
• [Fitting an Elastic Net with a precomputed Gram Matrix and Weighted Samples](The https://scikit-learn.org/stable/auto_examples/linear_model/plot_elastic_net_precomputed_gram_matrix_with_weighted_samples.html#sphx-glr-auto-examples-linear-model-plot-elastic-net-precomputed-gram-matrix-with-weighted-samples-py) can be removed and the content put in the docstring example section. (edit by @lorentzenchr)
• Gradient Boosting Out-of-Bag estimates and OOB Errors for Random Forests can be merged together (edit by @lorentzenchr)
• Two-class AdaBoost and [https://scikit-learn.org/stable/auto_examples/ensemble/plot_adaboost_multiclass.html](Multi-class AdaBoosted Decision Trees) can be merged together (edit by @lorentzenchr)

Please submit one PR per case.

Working on the example Plot randomly generated classification dataset.

[ ] Train error vs Test error example is really similar to Plotting Validation Curves example.

I'd like to work on this part, if it's available.

Edit: Merged.

@virchan It was merged

@glemaitre @ArturoAmorQ I would like to take Comparison of Calibration of Classifiers and Probability Calibration curves. Can I work on PR?

Those two examples in particular are pretty nice, not sure if we need to do anything about those ones?

@raviteja-ganta maybe you can merge Comparison of LDA and PCA 2D projection of Iris dataset and Incremental PCA, which are quite similar.

@ArturoAmorQ Looks like both are different. First one tries to compare LDA with PCA while second one(Incremental PCA) tries to compare between PCA and incremental PCA.