[Doc] Small fixes (fairlearn#1354)

* Add a type hint

* Silence some warnings

* Remove duplicate .bib entry

* More DF fixes

* Update for autogen file

* Get column type right

* Experimenting with some other doc changes

* Undo annotation changes

* blacken

.gitignore

@@ -66,6 +66,7 @@ instance/
 # Sphinx documentation
 docs/_build/
 docs/auto_examples/
+docs/sg_execution_times.rst
 
 # PyBuilder
 target/

docs/refs.bib

@@ -846,11 +846,3 @@ @misc{sykes2020boston
   title={- B 1000(Bk - 0.63)^2 where Bk is the proportion of blacks by town},
   url={https://github.com/scikit-learn/scikit-learn/issues/16155}
 }
-
-@misc{carlisle2019racist,
-  author={M Carlisle},
-  title={racist data destruction?},
-  publisher={Medium},
-  year={2019},
-  url={https://medium.com/@docintangible/racist-data-destruction-113e3eff54a8}
-}

examples/plot_grid_search_census.py

@@ -189,7 +189,7 @@ def classifier(X):
     disparity = DemographicParity()
     disparity.load_data(X_train, pd.Series(Y_train), sensitive_features=A_train)
 
-    errors.append(error.gamma(classifier)[0])
+    errors.append(error.gamma(classifier).iloc[0])
     disparities.append(disparity.gamma(classifier).max())
 
 all_results = pd.DataFrame(

examples/plot_intersectional_bias.py

@@ -1212,7 +1212,7 @@ def intersectionalf(truelabels, predictions):
 test["Diagnosis"] = test["Diagnosis"].astype(int)
 
 # Create new intersect column for plotting
-test["intersect"] = np.nan
+test["intersect"] = ""
 test.loc[(test["Sex"] == "Male") & (test["Race"] == "White"), "intersect"] = "WhiteM"
 test.loc[(test["Sex"] == "Male") & (test["Race"] == "Black"), "intersect"] = "BlackM"
 test.loc[(test["Sex"] == "Male") & (test["Race"] == "Hispanic"), "intersect"] = "HispanicM"

fairlearn/metrics/_plotter.py

@@ -263,16 +263,16 @@ def plot_metric_frame(
             for i in range(len(temp_axs[j].patches)):
                 temp_axs[j].text(
                     i,
-                    df_all_bounds[metric][i][0] - 0.05 * y_range,
-                    round(df_all_bounds[metric][i][0], ci_labels_precision),
+                    df_all_bounds[metric].iloc[i][0] - 0.05 * y_range,
+                    round(df_all_bounds[metric].iloc[i][0], ci_labels_precision),
                     fontsize=ci_labels_fontsize,
                     color=ci_labels_color,
                     ha=ci_labels_ha,
                 )
                 temp_axs[j].text(
                     i,
-                    df_all_bounds[metric][i][1] + 0.01 * y_range,
-                    round(df_all_bounds[metric][i][1], ci_labels_precision),
+                    df_all_bounds[metric].iloc[i][1] + 0.01 * y_range,
+                    round(df_all_bounds[metric].iloc[i][1], ci_labels_precision),
                     fontsize=ci_labels_fontsize,
                     color=ci_labels_color,
                     ha=ci_labels_ha,

fairlearn/reductions/_moments/error_rate.py

@@ -73,7 +73,7 @@ def load_data(self, X, y, *, sensitive_features, control_features=None):
         super().load_data(X, y_train, sensitive_features=sf_train)
         self.index = [_ALL]
 
-    def gamma(self, predictor):
+    def gamma(self, predictor) -> pd.Series:
         """Return the gamma values for the given predictor."""
         pred = predictor(self.X)
         if isinstance(pred, np.ndarray):