Rebase

Author: ArturoAmorQ

python_scripts/trees_sol_01.py

@@ -57,9 +57,9 @@
 #
 # ```{warning}
 # At this time, it is not possible to use `response_method="predict_proba"` for
-# multiclass problems. This is a planned feature for a future version of
-# scikit-learn. In the mean time, you can use `response_method="predict"`
-# instead.
+# multiclass problems on a single plot. This is a planned feature for a future
+# version of scikit-learn. In the mean time, you can use
+# `response_method="predict"` instead.
 # ```
 
 # %%
@@ -140,13 +140,15 @@
 # except that for a K-class problem you have K probability outputs for each
 # data point. Visualizing all these on a single plot can quickly become tricky
 # to interpret. It is then common to instead produce K separate plots, one for
-# each class, in a one-vs-rest (or one-vs-all) fashion.
+# each class, in a one-vs-rest (or one-vs-all) fashion. This can be achieved by
+# calling `DecisionBoundaryDisplay` several times, once for each class, and
+# passing the `class_of_interest` parameter to the function.
 #
-# For example, in the plot below, the first plot on the left shows in yellow the
-# certainty on classifying a data point as belonging to the "Adelie" class. In
-# the same plot, the spectre from green to purple represents the certainty of
-# **not** belonging to the "Adelie" class. The same logic applies to the other
-# plots in the figure.
+# For example, in the plot below, the first plot on the left shows the
+# certainty of classifying a data point as belonging to the "Adelie" class. The
+# darker the color, the more certain the model is that a given point in the
+# feature space belongs to a given class the predictions. The same logic
+# applies to the other plots in the figure.
 
 # %% tags=["solution"]
 # import numpy as np
@@ -166,6 +168,7 @@
         ax=ax,
         vmin=0,
         vmax=1,
+        cmap="Blues",
     )
     ax.scatter(
         data_test["Culmen Length (mm)"].loc[target_test == class_of_interest],
@@ -179,24 +182,17 @@
         ax.set_ylabel("Culmen Depth (mm)")
 
 ax = plt.axes([0.15, 0.14, 0.7, 0.05])
-plt.colorbar(
-    cm.ScalarMappable(cmap="viridis"), cax=ax, orientation="horizontal"
-)
-_ = ax.set_title("Class probability")
+plt.colorbar(cm.ScalarMappable(cmap="Blues"), cax=ax, orientation="horizontal")
+_ = ax.set_title("Predicted class membership probability")
 
 # %% [markdown] tags=["solution"]
+#
 # ```{note}
-# You may have noticed that we are no longer using a diverging colormap. Indeed,
-# the chance level for a one-vs-rest binarization of the multi-class
-# classification problem is almost never at predicted probability of 0.5. So
-# using a colormap with a neutral white at 0.5 might give a false impression on
-# the certainty.
+# You may notice that we do not use a diverging colormap (2 color gradients with
+# white in the middle). Indeed, in a multiclass setting, 0.5 is not a
+# meaningful value, hence using white as the center of the colormap is not
+# appropriate. Instead, we use a sequential colormap, where the color intensity
+# indicates the certainty of the classification. The darker the color, the more
+# certain the model is that a given point in the feature space belongs to a
+# given class.
 # ```
-#
-# Since scikit-learn v1.4, `DecisionBoundaryDisplay` supports a
-# `class_of_interest` parameter that allows in particular for a visualization of
-# `predict_proba` in multi-class settings.
-#
-# We also plan to make it possible to visualize the `predict_proba` values for
-# the class with the maximum predicted probability (without having to pass a
-# given a fixed `class_of_interest` value).