Add true_impact_distance to cross validation output

ctapipe/reco/sklearn.py

@@ -894,6 +894,7 @@ def __call__(self, telescope_type, table):
                         "prediction": cv_prediction,
                         "truth": truth,
                         "true_energy": test["true_energy"],
+                        "true_impact_distance": test["true_impact_distance"],
                     }
                 )
             )

ctapipe/tools/train_disp_reconstructor.py

@@ -119,6 +119,7 @@ def start(self):
             self.log.info("Loading events for %s", tel_type)
             feature_names = self.models.features + [
                 "true_energy",
+                "true_impact_distance",
                 "subarray_pointing_lat",
                 "subarray_pointing_lon",
                 "true_alt",
@@ -138,7 +139,10 @@ def start(self):
                 n_events=self.n_events.tel[tel_type],
             )
             table[self.models.target] = self._get_true_disp(table)
-            table = table[self.models.features + [self.models.target, "true_energy"]]
+            table = table[
+                self.models.features
+                + [self.models.target, "true_energy", "true_impact_distance"]
+            ]
 
             self.log.info("Train models on %s events", len(table))
             self.cross_validate(tel_type, table)

ctapipe/tools/train_energy_regressor.py

@@ -110,7 +110,10 @@ def start(self):
         self.log.info("Training models for %d types", len(types))
         for tel_type in types:
             self.log.info("Loading events for %s", tel_type)
-            feature_names = self.regressor.features + [self.regressor.target]
+            feature_names = self.regressor.features + [
+                self.regressor.target,
+                "true_impact_distance",
+            ]
             table = read_training_events(
                 loader=self.loader,
                 chunk_size=self.chunk_size,

ctapipe/tools/train_particle_classifier.py

@@ -174,6 +174,7 @@ def _read_input_data(self, tel_type):
         feature_names = self.classifier.features + [
             self.classifier.target,
             "true_energy",
+            "true_impact_distance",
         ]
         signal = read_training_events(
             loader=self.signal_loader,

docs/changes/2468.optimization.rst

@@ -0,0 +1 @@
+Add ``true_impact_distance`` to the output of ``CrossValidator``.