FAI-882: Add kwargs to explainers (#113)

* changed non-critical args to kwargs

* fixed missing bullet in lime kwargs docs

Author: RobGeada

src/trustyai/explainers/lime.py

@@ -194,7 +194,6 @@ def _get_bokeh_plot_dict(self):
         return plot_dict
 
 
-# pylint: disable=too-many-arguments
 class LimeExplainer:
     """*"Which features were most important to the results?"*
 
@@ -203,47 +202,49 @@ class LimeExplainer:
     feature that describe how strongly said feature contributed to the model's output.
     """
 
-    def __init__(
-        self,
-        perturbations=1,
-        seed=0,
-        samples=10,
-        penalise_sparse_balance=True,
-        track_counterfactuals=False,
-        normalise_weights=False,
-        use_wlr_model=True,
-        **kwargs,
-    ):
+    def __init__(self, samples=10, **kwargs):
         """Initialize the :class:`LimeExplainer`.
 
         Parameters
         ----------
-        perturbations: int
-            The starting number of feature perturbations within the explanation process.
-        seed: int
-            The random seed to be used.
         samples: int
             Number of samples to be generated for the local linear model training.
-        penalise_sparse_balance : bool
-            Whether to penalise features that are likely to produce linearly inseparable outputs.
-            This can improve the efficacy and interpretability of the outputted saliencies.
-        normalise_weights : bool
-            Whether to normalise the saliencies generated by LIME. If selected, saliencies will be
-            normalized between 0 and 1.
+
+        Keyword Arguments:
+            * penalise_sparse_balance : bool
+                (default=True) Whether to penalise features that are likely to produce linearly
+                inseparable outputs. This can improve the efficacy and interpretability of the
+                outputted saliencies.
+            * normalise_weights : bool
+                (default=False) Whether to normalise the saliencies generated by LIME. If selected,
+                saliencies will be normalized between 0 and 1.
+            * use_wlr_model : bool
+                (default=True) Whether to use a weighted linear regression as the LIME explanatory
+                model. If `false`, a multilayer perceptron is used, which generally has a slower
+                runtime,
+            * seed: int
+                (default=0) The random seed to be used.
+            * perturbations: int
+                (default=1) The starting number of feature perturbations within the explanation
+                process.
+            * trackCounterfactuals : bool
+                (default=False) Keep track of produced byproduct counterfactuals during LIME run.
         """
         self._jrandom = Random()
-        self._jrandom.setSeed(seed)
+        self._jrandom.setSeed(kwargs.get("seed", 0))
 
         self._lime_config = (
             LimeConfig()
-            .withNormalizeWeights(normalise_weights)
-            .withPerturbationContext(PerturbationContext(self._jrandom, perturbations))
+            .withNormalizeWeights(kwargs.get("normalise_weights", False))
+            .withPerturbationContext(
+                PerturbationContext(self._jrandom, kwargs.get("perturbations", 1))
+            )
             .withSamples(samples)
             .withEncodingParams(EncodingParams(0.07, 0.3))
             .withAdaptiveVariance(True)
-            .withPenalizeBalanceSparse(penalise_sparse_balance)
-            .withUseWLRLinearModel(use_wlr_model)
-            .withTrackCounterfactuals(track_counterfactuals)
+            .withPenalizeBalanceSparse(kwargs.get("penalise_sparse_balance", True))
+            .withUseWLRLinearModel(kwargs.get("use_wlr_model", True))
+            .withTrackCounterfactuals(kwargs.get("track_counterfactuals", False))
         )
 
         self._explainer = _LimeExplainer(self._lime_config)

src/trustyai/explainers/shap.py

@@ -1,6 +1,6 @@
 """Explainers.shap module"""
 # pylint: disable = import-error, too-few-public-methods, wrong-import-order, line-too-long,
-# pylint: disable = unused-argument, consider-using-f-string, invalid-name, too-many-arguments
+# pylint: disable = unused-argument, consider-using-f-string, invalid-name
 from typing import Dict, Optional, List, Union
 import matplotlib.pyplot as plt
 import matplotlib as mpl
@@ -434,11 +434,7 @@ class SHAPExplainer:
     def __init__(
         self,
         background: Union[np.ndarray, pd.DataFrame, List[PredictionInput]],
-        samples=None,
-        batch_size=20,
-        seed=0,
         link_type: Optional[_ShapConfig.LinkType] = None,
-        track_counterfactuals=False,
         **kwargs,
     ):
         r"""Initialize the :class:`SHAPxplainer`.
@@ -449,23 +445,26 @@ def __init__(
         or List[:class:`PredictionInput]
             The set of background datapoints as an array, dataframe of shape
             ``[n_datapoints, n_features]``, or list of TrustyAI PredictionInputs.
-        samples: int
-            The number of samples to use when computing SHAP values. Higher values will increase
-            explanation accuracy, at the  cost of runtime.
-        batch_size: int
-            The number of batches passed to the PredictionProvider at once. When using a
-            :class:`~Model` in the :func:`explain` function, this parameter has no effect. With an
-            :class:`~ArrowModel`, `batch_sizes` of around
-            :math:`\frac{2000}{\mathtt{len(background)}}` can produce significant
-            performance gains.
-        seed: int
-            The random seed to be used when generating explanations.
         link_type : :obj:`~_ShapConfig.LinkType`
             A choice of either ``trustyai.explainers._ShapConfig.LinkType.IDENTITY``
             or ``trustyai.explainers._ShapConfig.LinkType.LOGIT``. If the model output is a
             probability, choosing the ``LOGIT`` link will rescale explanations into log-odds units.
             Otherwise, choose ``IDENTITY``.
-
+        Keyword Arguments:
+            * samples: int
+                (default=None) The number of samples to use when computing SHAP values. Higher
+                values will increase explanation accuracy, at the  cost of runtime. If none,
+                samples will equal 2048 + 2*n_features
+            * seed: int
+                (default=0) The random seed to be used when generating explanations.
+            * batchSize: int
+                (default=20) The number of batches passed to the PredictionProvider at once.
+                When uusing :class:`~Model` with `arrow=False` this parameter has no effect.
+                If `arrow=True`, `batch_sizes` of around
+                :math:`\frac{2000}{\mathtt{len(background)}}` can produce significant
+                performance gains.
+            * trackCounterfactuals : bool
+                (default=False) Keep track of produced byproduct counterfactuals during SHAP run.
         Returns
         -------
         :class:`~SHAPResults`
@@ -474,7 +473,7 @@ def __init__(
         if not link_type:
             link_type = _ShapConfig.LinkType.IDENTITY
         self._jrandom = Random()
-        self._jrandom.setSeed(seed)
+        self._jrandom.setSeed(kwargs.get("seed", 0))
         perturbation_context = PerturbationContext(self._jrandom, 0)
 
         if isinstance(background, np.ndarray):
@@ -491,13 +490,13 @@ def __init__(
         self._configbuilder = (
             _ShapConfig.builder()
             .withLink(link_type)
-            .withBatchSize(batch_size)
+            .withBatchSize(kwargs.get("batch_size", 20))
             .withPC(perturbation_context)
             .withBackground(self.background)
-            .withTrackCounterfactuals(track_counterfactuals)
+            .withTrackCounterfactuals(kwargs.get("track_counterfactuals", False))
         )
-        if samples is not None:
-            self._configbuilder.withNSamples(JInt(samples))
+        if kwargs.get("samples") is not None:
+            self._configbuilder.withNSamples(JInt(kwargs["samples"]))
         self._config = self._configbuilder.build()
         self._explainer = _ShapKernelExplainer(self._config)