24424: Updates details docstrings for new "relevant_values" detail in both #react and #react_aggregate, MINOR

howso/client/base.py

@@ -1969,6 +1969,10 @@ def react(  # noqa: C901
                 Uses only the context features of the reacted case to determine that area.
                 Uses full calculations, which uses leave-one-out context features for
                 computations.
+            - relevant_values : bool or list of strings, optional
+                When true outputs a map of each context feature name to a list of relevant values for that
+                feature given the context. If a list of feature names, will only output relevant values for
+                each feature specified.
             - selected_prediction_stats : list, optional.
                 List of stats to output. When unspecified, returns all except the confusion matrix. Allowed values:
 
@@ -2263,15 +2267,6 @@ def react(  # noqa: C901
                 " following values - ['min', 'max', 'most_similar',]"
             )
 
-        if details is not None and 'robust_computation' in details:
-            details = dict(details)
-            details['robust_influences'] = details['robust_computation']
-            details['robust_residuals'] = details['robust_computation']
-            del details['robust_computation']
-            warnings.warn(
-                'The detail "robust_computation" is deprecated and will be '
-                'removed in a future release. Please use "robust_residuals" '
-                'and/or "robust_influences" instead.', DeprecationWarning)
 
         if details is not None and 'local_case_feature_residual_conviction_robust' in details:
             details = dict(details)
@@ -2752,7 +2747,7 @@ def _preprocess_generate_parameters(  # noqa: C901
                     target_context_values = context_values[i]
 
             if context_features and (
-                not target_context_values or
+                target_context_values is None or
                 not isinstance(target_context_values, Sized) or
                 len(target_context_values) != len(context_features)
             ):
@@ -3906,6 +3901,10 @@ def react_aggregate(  # noqa: C901
                 Compute accuracy contributions by scrambling each feature and
                 using the robust (power set/permutations) set of remaining
                 context features for each prediction.
+            - relevant_values : bool or list of strings, optional
+                When true outputs a map of each context feature name to a list of relevant values for that
+                feature given the context. If a list of feature names, will only output relevant values for
+                each feature specified.
             - action_condition : map of str -> any, optional
                 A condition map to select the action set, which is the collection of cases
                 reacted to while computing the requested metrics.
@@ -4917,19 +4916,19 @@ def set_auto_ablation_params(
         auto_ablation_enabled: bool = False,
         *,
         ablated_cases_distribution_batch_size: int = 100,
-        abs_threshold_map: AblationThresholdMap = None,
+        abs_threshold_map: t.Optional[AblationThresholdMap] = None,
         auto_ablation_influence_weight_entropy_threshold: float = 0.15,
         auto_ablation_weight_feature: str = ".case_weight",
         batch_size: int = 2_000,
         conviction_lower_threshold: t.Optional[float] = None,
         conviction_upper_threshold: t.Optional[float] = None,
-        delta_threshold_map: AblationThresholdMap = None,
+        delta_threshold_map: t.Optional[AblationThresholdMap] = None,
         exact_prediction_features: t.Optional[Collection[str]] = None,
         influence_weight_entropy_sample_size: int = 2_000,
         min_num_cases: int = 10_000,
         max_num_cases: int = 200_000,
         reduce_data_influence_weight_entropy_threshold: float = 0.6,
-        rel_threshold_map: AblationThresholdMap = None,
+        rel_threshold_map: t.Optional[AblationThresholdMap] = None,
         relative_prediction_threshold_map: t.Optional[Mapping[str, float]] = None,
         residual_prediction_features: t.Optional[Collection[str]] = None,
         tolerance_prediction_threshold_map: t.Optional[Mapping[str, tuple[float, float]]] = None,

howso/client/schemas/reaction.py

@@ -89,7 +89,7 @@ class Reaction(abc.MutableMapping):
         "observational_errors",
         "outlying_feature_values",
         "prediction_stats",
-        "robust_influences",
+        "relevant_values",
         "similarity_conviction",
         # react_series-only details
         "aggregated_categorical_action_probabilities",

howso/engine/trainee.py

@@ -760,19 +760,19 @@ def set_auto_ablation_params(
         auto_ablation_enabled: bool = False,
         *,
         ablated_cases_distribution_batch_size: int = 100,
-        abs_threshold_map: AblationThresholdMap = None,
+        abs_threshold_map: t.Optional[AblationThresholdMap] = None,
         auto_ablation_influence_weight_entropy_threshold: float = 0.15,
         auto_ablation_weight_feature: str = ".case_weight",
         batch_size: int = 2_000,
         conviction_lower_threshold: t.Optional[float] = None,
         conviction_upper_threshold: t.Optional[float] = None,
-        delta_threshold_map: AblationThresholdMap = None,
+        delta_threshold_map: t.Optional[AblationThresholdMap] = None,
         exact_prediction_features: t.Optional[Collection[str]] = None,
         influence_weight_entropy_sample_size: int = 2_000,
         min_num_cases: int = 10_000,
         max_num_cases: int = 200_000,
         reduce_data_influence_weight_entropy_threshold: float = 0.6,
-        rel_threshold_map: AblationThresholdMap = None,
+        rel_threshold_map: t.Optional[AblationThresholdMap] = None,
         relative_prediction_threshold_map: t.Optional[Mapping[str, float]] = None,
         residual_prediction_features: t.Optional[Collection[str]] = None,
         tolerance_prediction_threshold_map: t.Optional[Mapping[str, tuple[float, float]]] = None,
@@ -1607,6 +1607,10 @@ def react(
                 Uses only the context features of the reacted case to determine that area.
                 Uses full calculations, which uses leave-one-out context features for
                 computations.
+            - relevant_values : bool or list of strings, optional
+                When true outputs a map of each context feature name to a list of relevant values for that
+                feature given the context. If a list of feature names, will only output relevant values for
+                each feature specified.
             - selected_prediction_stats : list[Prediction_Stats], optional.
                 List of stats to output. When unspecified, returns all except the confusion matrix. Allowed values:
 
@@ -3613,6 +3617,10 @@ def react_aggregate(
                 "selected_prediction_stats" parameter in the `details` parameter.
                 Uses full calculations, which uses leave-one-out for features for
                 computations.
+            - relevant_values : bool or list of strings, optional
+                When true outputs a map of each context feature name to a list of relevant values for that
+                feature given the context. If a list of feature names, will only output relevant values for
+                each feature specified.
             - selected_prediction_stats : list, optional
                 List of stats to output. When unspecified, returns all except the confusion matrix. Allowed values:
 

howso/scikit/scikit.py

@@ -871,6 +871,10 @@ def describe_prediction(self, X: np.ndarray, details: t.Optional[Mapping] = None
                 Uses only the context features of the reacted case to determine that area.
                 Uses full calculations, which uses leave-one-out context features for
                 computations.
+            - relevant_values : bool or list of strings, optional
+                When true outputs a map of each context feature name to a list of relevant values for that
+                feature given the context. If a list of feature names, will only output relevant values for
+                each feature specified.
             - selected_prediction_stats : list, optional
                 List of stats to output. When unspecified, returns all except the confusion matrix. Allowed values:
 

howso/utilities/internals.py

@@ -913,7 +913,7 @@ def to_pandas_datetime_format(f: str):
 
 def fix_feature_value_keys(
     input_dict: dict[str, t.Any],
-    feature_attributes: Mapping[t.Hashable, Mapping],
+    feature_attributes: Mapping[str, Mapping],
     feature_name: str
 ) -> dict[str | float | int, t.Any]:
     """
@@ -926,7 +926,7 @@ def fix_feature_value_keys(
     ----------
     input_dict : dict[str, Any]
         The mapping with feature values as keys that may need fixing.
-    feature_attributes : Mapping[Hashable, Mapping]
+    feature_attributes : Mapping[str, Mapping]
         The feature attributes of the data.
     feature_name : str
         The name of the feature whose feature values make the keys of the dict.
@@ -951,9 +951,9 @@ def fix_feature_value_keys(
     return output_dict
 
 def update_caps_maps(
-    caps_maps: list[dict[dict[str, float]]],
-    feature_attributes: Mapping[t.Hashable, Mapping]
-) -> list[dict[dict[str | int | float, float]]]:
+    caps_maps: list[dict[str, dict[str, float]]],
+    feature_attributes: Mapping[str, Mapping]
+) -> list[dict[str, dict[str | int | float, float]]]:
     """
     Cleans up misformatted keys from non-string nominal feature's CAP maps.
 
@@ -962,14 +962,14 @@ def update_caps_maps(
 
     Parameters
     ----------
-    caps_maps : list[dict[dict[str, float]]]
+    caps_maps : list[dict[str, dict[str, float]]]
         The list of CAP maps.
     feature_attributes : Mapping[str, Mapping]
         The feature attributes of the data.
 
     Returns
     -------
-    list[dict[dict[str | int | float, float]]]
+    list[dict[str, dict[str | int | float, float]]]
         The updated list of CAP maps with cleaned up feature values as keys.
     """
     updated_caps_maps = []
@@ -987,8 +987,8 @@ def update_caps_maps(
     return updated_caps_maps
 
 def update_confusion_matrix(
-    confusion_matrix: dict[str, dict | float],
-    feature_attributes: Mapping[t.Hashable, Mapping]
+    confusion_matrix: dict[str, dict[str, float | dict[str, t.Any]]],
+    feature_attributes: Mapping[str, Mapping]
 ) -> dict[str, t.Any]:
     """
     Cleans up misformatted keys from non-string nominal feature's confusion matrices.
@@ -998,9 +998,9 @@ def update_confusion_matrix(
 
     Parameters
     ----------
-    confusion_matrix : dict[str, dict | float]
+    confusion_matrix : dict[str, dict[str, float | dict[str, t.Any]]]
         The mapping that defines the confusion matrix.
-    feature_attributes : Mapping[Hashable, Mapping]
+    feature_attributes : Mapping[str, Mapping]
         The feature attributes of the data.
 
     Returns