FutureWarnings in the documentation

[ArturoAmorQ]

Describe the issue linked to the documentation

Some FutureWarnings are present in the dev documentation and need to be fixed.
Here is a list:

• bicluster/plot_bicluster_newsgroups.html
• cluster/plot_bisect_kmeans.html DOC remove FutureWarning in cluster/plot_bisect_kmeans.py #24891
• decomposition/plot_ica_blind_source_separation.html DOC Fix FutureWarning in decomposition/plot_ica_blind_source_separation.html #24908
• ensemble/plot_gradient_boosting_categorical.html DOC Fix FutureWarning in ensemble/plot_gradient_boosting_categorical example #24902
• cluster/plot_color_quantization.html DOC remove FutureWarning in plot_colour_quantization example #24893
• ensemble/plot_stack_predictors.html DOC remove FutureWarnings from plot_stack_predictors #24918
• inspection/plot_linear_model_coefficient_interpretation.html DOC Fix FutureWarning in inspection/plot_linear_model_coefficient_interpretation #24920
• cross_decomposition/plot_compare_cross_decomposition.html
• cluster/plot_cluster_comparison.html DOC Fix FutureWarning in cluster/plot_cluster_comparison.html #24927
• model_selection/plot_successive_halving_heatmap.html DOC Fix FutureWarning in plot_successive_halving_heatmap.html #24968
• manifold/plot_compare_methods.html DOC Fix FutureWarning in manifold/plot_compare_methods.py #24909
• cluster/plot_kmeans_assumptions.html DOC Fix FutureWarning in cluster/plot_kmeans_assumptions.html #24928
• decomposition/plot_faces_decomposition.html DOC Fix FutureWarning in decomposition/plot_faces_decomposition.html #24936
• ensemble/plot_gradient_boosting_oob.html DOC Fix FutureWarning in ensemble/plot_gradient_boosting_oob.py #24948
• ensemble/plot_gradient_boosting_regression.html DOC Fix FutureWarning in ensemble/plot_gradient_boosting_regression.html #24946
• ensemble/plot_gradient_boosting_regularization.html DOC Fix FutureWarning in ensemble/plot_gradient_boosting_regularization.html #24960
• manifold/plot_manifold_sphere.html DOC remove FutureWarnings in plot_manifold_sphere example #24941
• manifold/plot_lle_digits.html DOC Fix FutureWarning in manifold/plot_lle_digits.html #24922
• manifold/plot_mds.html DOCremove FutureWarning in manifold/plot_mds.html #24940
• cluster/plot_dict_face_patches.html DOC Fix FutureWarning in 'cluster/plot_dict_face_patches.html' #24910
• release_highlights/plot_release_highlights_0_23_0.html DOC remove FutureWarnings for plot_release_highlights_0_23_0.py #24937
• release_highlights/plot_release_highlights_1_1_0.html DOC remove FutureWarnings from release 1.1.0 #24939
• model_selection/plot_grid_search_text_feature_extraction.html DOC Fix FutureWarning in model_selection/plot_grid_search_text_feature_extraction.html #24961
• cluster/plot_kmeans_silhouette_analysis.html DOC Fix FutureWarning in cluster/plot_kmeans_silhouette_analysis.html #24929
• applications/plot_cyclical_feature_engineering.html DOC Fix FutureWarning in 'applications/plot_cyclical_feature_engineering.html' #24892

Contributors willing to address this issue, please offer one example per pull request.

Thanks for your help!

Suggest a potential alternative/fix

No response

[Ti-Ion]

Hi, may I work on applications/plot_cyclical_feature_engineering.html?

[ArturoAmorQ]

Hi, may I work on applications/plot_cyclical_feature_engineering.html?

Sure, feel free to open a PR mentioning this issue :)

[GeorgiaMayDay]

Sorry, this is my first time contributing to sci-learn. When you say fix, should we make it so the future warnings are accurate and change the defaults? Or instead address the TODO's in the code and then remove the warnings?

So, in the case of https://scikit-learn.org/dev/auto_examples/cluster/plot_color_quantization.html - should I:

• Change default_n_init to 10
  or
• Fix it so default_n_init is set by auto

[ArturoAmorQ]

Sorry, this is my first time contributing to sci-learn. When you say fix, should we make it so the future warnings are accurate and change the defaults? Or instead address the TODO's in the code and then remove the warnings?

You have to change the code of each given example so that the warning is no longer raised. In the case of cluster/plot_color_quantization.html the message says

Set the value of `n_init` explicitly to suppress the warning.

So the thing to do is to set n_init=5 explicitly, i.e.

print("Fitting model on a small sub-sample of the data")
t0 = time()
image_array_sample = shuffle(image_array, random_state=0, n_samples=1_000)
- kmeans = KMeans(n_clusters=n_colors, random_state=0).fit(image_array_sample)
+ kmeans = KMeans(n_clusters=n_colors, n_init="auto", random_state=0).fit(image_array_sample)
print(f"done in {time() - t0:0.3f}s.")

[aditya-anulekh]

Submitted a PR to fix FutureWarning in examples/cluster/plot_bisect_kmeans.py

[DhanshreeA]

Working on ensemble/plot_gradient_boosting_categorical.html

[Ti-Ion]

Working on cluster/plot_dict_face_patches.html

[SarahRemus]

Working on inspection/plot_linear_model_coefficient_interpretation.html

[Reddragonemperor58]

Working on cross_decomposition/plot_compare_cross_decomposition.html

[Reddragonemperor58]

Hi @ArturoAmorQ I am working on cross_decomposition and I found the future warning i.e., "_/srv/conda/envs/notebook/lib/python3.9/site-packages/sklearn/cross_decomposition/pls.py:503: FutureWarning: The attribute coef_ will be transposed in version 1.3 to be consistent with other linear models in scikit-learn. Currently, coef_ has a shape of (n_features, n_targets) and in the future, it will have a shape of (n_targets, n_features).
warnings.warn("

This warning cannot be suppressed by changing anything from the code in documentation, I think we need to edit the main file of 'coef_' from here: https://github.com/scikit-learn/scikit-learn/blob/main/sklearn/cross_decomposition/_pls.py

In line number 504 we need to make changes. Or we can just suppress it by using simplefilter.
Could you confirm it?

[ArturoAmorQ]

This warning cannot be suppressed by changing anything from the code in documentation, I think we need to edit the main file of 'coef_' from here: main/sklearn/cross_decomposition/_pls.py

In line number 504 we need to make changes. Or we can just suppress it by using simplefilter. Could you confirm it?

Changing the coef_ function is a TODO for version 1.3, so this should not be changed for the moment. I think it's better not to change the example either.

We will have to keep this particular example in mind. Thanks for the report @Reddragonemperor58 !

[Reddragonemperor58]

Welcome, you could remove it from the list.

[jpangas]

Working on decomposition/plot_faces_decomposition.html

[Reddragonemperor58]

working on model_selection/plot_successive_halving_heatmap.html

[Reddragonemperor58]

Hey @ArturoAmorQ for model_selection/plot_successive_halving_heatmap.html the warning raised is "In a future version, the Index constructor will not infer numeric dtypes when passed object-dtype sequences (matching Series behavior)" is it okay to suppress the warning using ignore as they did it in pandas issue ?

[edenies]

Working on release_highlights/plot_release_highlights_0_23_0.html

[Reddragonemperor58]

working on manifold/plot_manifold_sphere.html

[edenies]

Working on release_highlights/plot_release_highlights_1_1_0.html

[Reddragonemperor58]

working on ensemble/plot_gradient_boosting_regularization.html

[isaac-chung]

Working on model_selection/plot_grid_search_text_feature_extraction.html

[ArturoAmorQ]

Thanks to everyone who contributed on this issue!