FAI-926: Implement custom counterfactual goal criteria (#140)

* Add initial GoalCriteria
* Refactor wrapped functions abstract class
* Refactor output casting
* Add support for numpy CF criteria
* Add test for missing goal and criteria
* Fix formatting
* Fix feature domain import

Author: ruivieira

src/trustyai/explainers/counterfactuals.py

@@ -18,9 +18,9 @@
     counterfactual_prediction,
     PredictionInput,
     Model,
+    GoalCriteria,
 )
 
-
 from trustyai.utils.data_conversions import (
     prediction_object_to_numpy,
     prediction_object_to_pandas,
@@ -184,12 +184,13 @@ def __init__(self, steps=10_000):
     def explain(
         self,
         inputs: OneInputUnionType,
-        goal: OneOutputUnionType,
         model: Union[PredictionProvider, Model],
+        goal: Optional[OneOutputUnionType] = None,
         feature_domains: List[FeatureDomain] = None,
         data_distribution: Optional[DataDistribution] = None,
         uuid: Optional[_uuid.UUID] = None,
         timeout: Optional[float] = None,
+        criteria: Optional[GoalCriteria] = None,
     ) -> CounterfactualResult:
         r"""Request for a counterfactual explanation given a list of features, goals and a
         :class:`~PredictionProvider`
@@ -217,7 +218,9 @@ def explain(
         uuid : Optional[:class:`_uuid.UUID`]
             The UUID to use during search.
         timeout : Optional[float]
-                The timeout time in seconds of the counterfactual explanation.
+            The timeout time in seconds of the counterfactual explanation.
+        criteria : Optional[:class:`GoalCriteria`]
+            An optional custom scoring function, wrapped as a :class:`GoalCriteria`.
 
         Returns
         -------
@@ -226,6 +229,9 @@ def explain(
         """
         feature_names = model.feature_names if isinstance(model, Model) else None
         output_names = model.output_names if isinstance(model, Model) else None
+        if goal is None and criteria is None:
+            raise ValueError("Either a goal or criteria must be provided.")
+
         _prediction = counterfactual_prediction(
             input_features=one_input_convert(
                 inputs, feature_names=feature_names, feature_domains=feature_domains
@@ -236,6 +242,7 @@ def explain(
             data_distribution=data_distribution,
             uuid=uuid,
             timeout=timeout,
+            criteria=criteria,
         )
 
         with Model.NonArrowTransmission(model):