From b471a7c4f2cb34707eb0e35b27ff2e178ea0aea0 Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Mon, 14 Nov 2022 15:06:04 +0800
Subject: [PATCH 01/15] Add global Shap values

---
 omnixai/explainers/tabular/agnostic/shap.py   | 65 +++++++++++++++----
 .../shap/test_shap_tabular_global.py          | 33 ++++++++++
 2 files changed, 84 insertions(+), 14 deletions(-)
 create mode 100644 omnixai/tests/explainers/shap/test_shap_tabular_global.py

diff --git a/omnixai/explainers/tabular/agnostic/shap.py b/omnixai/explainers/tabular/agnostic/shap.py
index 8afb78d1..e29abb36 100644
--- a/omnixai/explainers/tabular/agnostic/shap.py
+++ b/omnixai/explainers/tabular/agnostic/shap.py
@@ -13,7 +13,8 @@
 
 from ..base import TabularExplainer
 from ....data.tabular import Tabular
-from ....explanations.tabular.feature_importance import FeatureImportance
+from ....explanations.tabular.feature_importance import \
+    FeatureImportance, GlobalFeatureImportance
 
 
 class ShapTabular(TabularExplainer):
@@ -22,7 +23,7 @@ class ShapTabular(TabularExplainer):
     If using this explainer, please cite the original work: https://github.com/slundberg/shap.
     """
 
-    explanation_type = "local"
+    explanation_type = "both"
     alias = ["shap"]
 
     def __init__(
@@ -57,19 +58,34 @@ def __init__(
             kwargs["nsamples"] = 100
         self.background_data = shap.sample(self.data, nsamples=kwargs["nsamples"])
 
-    def explain(self, X, y=None, **kwargs) -> FeatureImportance:
-        """
-        Generates the feature-importance explanations for the input instances.
+    def _explain_global(self, **kwargs) -> GlobalFeatureImportance:
+        if "nsamples" not in kwargs:
+            kwargs["nsamples"] = 100
 
-        :param X: A batch of input instances. When ``X`` is `pd.DataFrame`
-            or `np.ndarray`, ``X`` will be converted into `Tabular` automatically.
-        :param y: A batch of labels to explain. For regression, ``y`` is ignored.
-            For classification, the top predicted label of each instance will be explained
-            when ``y = None``.
-        :param kwargs: Additional parameters for `shap.KernelExplainer.shap_values`,
-            e.g., ``nsamples`` -- the number of times to re-evaluate the model when explaining each prediction.
-        :return: The feature-importance explanations for all the input instances.
-        """
+        instances = self.background_data
+        explanations = GlobalFeatureImportance()
+        explainer = shap.KernelExplainer(
+            self.predict_fn, self.background_data,
+            link="logit" if self.mode == "classification" else "identity", **kwargs
+        )
+        shap_values = explainer.shap_values(instances, **kwargs)
+
+        if self.mode == "classification":
+            values = 0
+            for v in shap_values:
+                values += np.abs(v)
+            values /= len(shap_values)
+            shap_values = values
+
+        importance_scores = np.mean(np.abs(shap_values), axis=0)
+        explanations.add(
+            feature_names=self.feature_columns,
+            importance_scores=importance_scores,
+            sort=True
+        )
+        return explanations
+
+    def _explain_local(self, X, y=None, **kwargs) -> FeatureImportance:
         X = self._to_tabular(X).remove_target_column()
         explanations = FeatureImportance(self.mode)
         instances = self.transformer.transform(X)
@@ -120,6 +136,7 @@ def _predict(_x):
                     _y = np.tile(instance, (_x.shape[0], 1))
                     _y[:, self.valid_indices] = _x
                     return self.predict_fn(_y)
+
                 predict_function = _predict
                 test_x = instance[self.valid_indices]
 
@@ -148,6 +165,26 @@ def _predict(_x):
                 )
         return explanations
 
+    def explain(self, X=None, y=None, **kwargs):
+        """
+        Generates the local SHAP explanations for the input instances or global SHAP explanations.
+
+        :param X: A batch of input instances. When ``X`` is `pd.DataFrame`
+            or `np.ndarray`, ``X`` will be converted into `Tabular` automatically. If X is None,
+            it will compute global SHAP values, i.e., the mean of the absolute SHAP value for
+            each feature.
+        :param y: A batch of labels to explain. For regression, ``y`` is ignored.
+            For classification, the top predicted label of each instance will be explained
+            when ``y = None``.
+        :param kwargs: Additional parameters for `shap.KernelExplainer.shap_values`,
+            e.g., ``nsamples`` -- the number of times to re-evaluate the model when explaining each prediction.
+        :return: The feature-importance explanations.
+        """
+        if X is not None:
+            return self._explain_local(X=X, y=y, **kwargs)
+        else:
+            return self._explain_global(**kwargs)
+
     def save(
             self,
             directory: str,
diff --git a/omnixai/tests/explainers/shap/test_shap_tabular_global.py b/omnixai/tests/explainers/shap/test_shap_tabular_global.py
new file mode 100644
index 00000000..9b1d3938
--- /dev/null
+++ b/omnixai/tests/explainers/shap/test_shap_tabular_global.py
@@ -0,0 +1,33 @@
+#
+# Copyright (c) 2022 salesforce.com, inc.
+# All rights reserved.
+# SPDX-License-Identifier: BSD-3-Clause
+# For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/BSD-3-Clause
+#
+import os
+import unittest
+import pprint
+from omnixai.utils.misc import set_random_seed
+from omnixai.explainers.tabular import ShapTabular
+from omnixai.tests.explainers.tasks import TabularClassification
+
+
+class TestShapTabular(unittest.TestCase):
+    def test(self):
+        base_folder = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..")
+        task = TabularClassification(base_folder).train_adult(num_training_samples=2000)
+        predict_function = lambda z: task.model.predict_proba(task.transform.transform(z))
+
+        set_random_seed()
+        explainer = ShapTabular(
+            training_data=task.train_data,
+            predict_function=predict_function,
+            ignored_features=None,
+            nsamples=100
+        )
+        explanations = explainer.explain()
+        pprint.pprint(explanations.get_explanations())
+
+
+if __name__ == "__main__":
+    unittest.main()

From 31aafb7c74060810bdce36ea4368cc9ab8055c7a Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Mon, 14 Nov 2022 15:06:25 +0800
Subject: [PATCH 02/15] Update setup.py

---
 setup.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/setup.py b/setup.py
index e553bc10..eaa34d15 100644
--- a/setup.py
+++ b/setup.py
@@ -16,7 +16,7 @@
 
 setup(
     name="omnixai",
-    version="1.2.2",
+    version="1.2.3",
     author="Wenzhuo Yang, Hung Le, Tanmay Shivprasad Laud, Silvio Savarese, Steven C.H. Hoi",
     description="OmniXAI: An Explainable AI Toolbox",
     long_description=open("README.md", "r", encoding="utf-8").read(),

From fc924ae862de0d8124d6de7af9fd715e0d791b0b Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Mon, 14 Nov 2022 20:33:59 +0800
Subject: [PATCH 03/15] Add a KNN-based counterfactual explainer

---
 omnixai/explainers/tabular/__init__.py        |   2 +
 .../explainers/tabular/counterfactual/knn.py  | 112 ++++++++++++++++++
 .../tabular/counterfactual/mace/retrieval.py  |  18 ++-
 omnixai/tests/explainers/knn/test_ce_knn.py   |  39 ++++++
 .../tests/visualization/dashboard_tabular.py  |   2 +-
 5 files changed, 168 insertions(+), 5 deletions(-)
 create mode 100644 omnixai/explainers/tabular/counterfactual/knn.py
 create mode 100644 omnixai/tests/explainers/knn/test_ce_knn.py

diff --git a/omnixai/explainers/tabular/__init__.py b/omnixai/explainers/tabular/__init__.py
index 0a1dcdb6..31737c46 100644
--- a/omnixai/explainers/tabular/__init__.py
+++ b/omnixai/explainers/tabular/__init__.py
@@ -13,6 +13,7 @@
 from .agnostic.L2X.l2x import L2XTabular
 from .counterfactual.mace.mace import MACEExplainer
 from .counterfactual.ce import CounterfactualExplainer
+from .counterfactual.knn import KNNCounterfactualExplainer
 from .specific.ig import IntegratedGradientTabular
 from .specific.linear import LinearRegression
 from .specific.linear import LogisticRegression
@@ -31,6 +32,7 @@
     "L2XTabular",
     "MACEExplainer",
     "CounterfactualExplainer",
+    "KNNCounterfactualExplainer",
     "LinearRegression",
     "LogisticRegression",
     "TreeRegressor",
diff --git a/omnixai/explainers/tabular/counterfactual/knn.py b/omnixai/explainers/tabular/counterfactual/knn.py
new file mode 100644
index 00000000..0c07500c
--- /dev/null
+++ b/omnixai/explainers/tabular/counterfactual/knn.py
@@ -0,0 +1,112 @@
+#
+# Copyright (c) 2022 salesforce.com, inc.
+# All rights reserved.
+# SPDX-License-Identifier: BSD-3-Clause
+# For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/BSD-3-Clause
+#
+"""
+The KNN-based counterfactual explainer for tabular data.
+"""
+import numpy as np
+import pandas as pd
+from typing import List, Callable, Union
+
+from ..base import ExplainerBase
+from ...tabular.base import TabularExplainerMixin
+from ....data.tabular import Tabular
+from ....explanations.tabular.counterfactual import CFExplanation
+
+from .mace.retrieval import CFRetrieval
+from .mace.diversify import DiversityModule
+
+
+class KNNCounterfactualExplainer(ExplainerBase, TabularExplainerMixin):
+    """
+    The counterfactual explainer for tabular data based on KNN search. Given a query instance,
+    it finds the instances in the training dataset that are close to the query with the desired label.
+    """
+    explanation_type = "local"
+    alias = ["ce_knn", "knn_ce"]
+
+    def __init__(
+            self,
+            training_data: Tabular,
+            predict_function: Callable,
+            mode: str = "classification",
+            **kwargs,
+    ):
+        """
+        :param training_data: The data used to initialize a MACE explainer. ``training_data``
+            can be the training dataset for training the machine learning model.
+        :param predict_function: The prediction function corresponding to the model to explain.
+            The model should be a classifier, the outputs of the ``predict_function``
+            are the class probabilities.
+        :param mode: The task type can be `classification` only.
+        """
+        super().__init__()
+        assert mode == "classification", "MACE supports classification tasks only."
+        self.predict_function = predict_function
+
+        self.categorical_columns = training_data.categorical_columns
+        self.target_column = training_data.target_column
+        self.original_feature_columns = training_data.columns
+
+        self.recall = CFRetrieval(training_data, predict_function, None, **kwargs)
+        self.diversity = DiversityModule(training_data)
+
+    def explain(
+            self,
+            X: Tabular,
+            y: Union[List, np.ndarray] = None,
+            max_number_examples: int = 5,
+            **kwargs
+    ) -> CFExplanation:
+        """
+        Generates counterfactual explanations.
+
+        :param X: A batch of input instances. When ``X`` is `pd.DataFrame`
+            or `np.ndarray`, ``X`` will be converted into `Tabular` automatically.
+        :param y: A batch of the desired labels, which should be different from the predicted labels of ``X``.
+            If ``y = None``, the desired labels will be the labels different from the predicted labels of ``X``.
+        :param max_number_examples: The maximum number of the generated counterfactual
+            examples per class for each input instance.
+        :return: A CFExplanation object containing the generated explanations.
+        """
+        if y is not None:
+            assert len(y) == X.shape[0], (
+                f"The length of `y` should equal the number of instances in `X`, " f"got {len(y)} != {X.shape[0]}"
+            )
+
+        X = self._to_tabular(X).remove_target_column()
+        scores = self.predict_function(X)
+        labels = np.argmax(scores, axis=1)
+        num_classes = scores.shape[1]
+
+        explanations = CFExplanation()
+        for i in range(X.shape[0]):
+            x = X.iloc(i)
+            label = int(labels[i])
+            if y is None or y[i] == label:
+                desired_labels = [z for z in range(num_classes) if z != label]
+            else:
+                desired_labels = [int(y[i])]
+
+            all_cfs = []
+            for desired_label in desired_labels:
+                df, _ = self.recall.get_nn_samples(x, desired_label)
+                examples = Tabular(df, categorical_columns=x.categorical_columns)
+                cfs = self.diversity.get_diverse_cfs(
+                    self.predict_function, x, examples,
+                    oracle_function=lambda _s: int(desired_label == np.argmax(_s)),
+                    desired_label=desired_label, k=max_number_examples
+                )
+                cfs_df = cfs.to_pd()
+                if x.continuous_columns:
+                    cfs_df = cfs_df.astype({c: float for c in x.continuous_columns})
+                cfs_df["label"] = desired_label
+                all_cfs.append(cfs_df)
+
+            instance_df = x.to_pd()
+            instance_df["label"] = label
+            explanations.add(query=instance_df, cfs=pd.concat(all_cfs) if len(all_cfs) > 0 else None)
+        return explanations
diff --git a/omnixai/explainers/tabular/counterfactual/mace/retrieval.py b/omnixai/explainers/tabular/counterfactual/mace/retrieval.py
index 3692e28f..6cc8a108 100644
--- a/omnixai/explainers/tabular/counterfactual/mace/retrieval.py
+++ b/omnixai/explainers/tabular/counterfactual/mace/retrieval.py
@@ -152,13 +152,13 @@ def _pick_top_columns(self, x: Tabular, candidate_features: Dict, desired_label:
         columns = [c for c, _ in column_scores][:top_k]
         return {f: v for f, v in candidate_features.items() if f in columns}
 
-    def get_cf_features(self, instance: Tabular, desired_label: int) -> (Dict, np.ndarray):
+    def get_nn_samples(self, instance: Tabular, desired_label: int) -> (pd.DataFrame, np.ndarray):
         """
-        Finds candidate features for generating counterfactual examples.
+        Finds nearest neighbor samples in a desired class.
 
         :param instance: The query instance.
         :param desired_label: The desired label.
-        :return: The candidate features and the indices of the nearest neighbors.
+        :return: The nearest neighbor samples and the corresponding indices.
         """
         assert isinstance(instance, Tabular), "Input ``instance`` should be an instance of Tabular."
         assert instance.shape[0] == 1, "Input ``instance`` can only contain one instance."
@@ -171,9 +171,19 @@ def get_cf_features(self, instance: Tabular, desired_label: int) -> (Dict, np.nd
             )
         )
         indices = self._knn_query(query, desired_label, self.num_neighbors)[0]
+        y = self.subset.iloc(indices).to_pd(copy=False)
+        return y, indices
 
+    def get_cf_features(self, instance: Tabular, desired_label: int) -> (Dict, np.ndarray):
+        """
+        Finds candidate features for generating counterfactual examples.
+
+        :param instance: The query instance.
+        :param desired_label: The desired label.
+        :return: The candidate features and the indices of the nearest neighbors.
+        """
         x = instance.to_pd(copy=False)
-        y = self.subset.iloc(indices).to_pd(copy=False)
+        y, indices = self.get_nn_samples(instance, desired_label)
         cate_candidates, cont_candidates = {}, {}
 
         # Categorical feature difference
diff --git a/omnixai/tests/explainers/knn/test_ce_knn.py b/omnixai/tests/explainers/knn/test_ce_knn.py
new file mode 100644
index 00000000..208e9954
--- /dev/null
+++ b/omnixai/tests/explainers/knn/test_ce_knn.py
@@ -0,0 +1,39 @@
+#
+# Copyright (c) 2022 salesforce.com, inc.
+# All rights reserved.
+# SPDX-License-Identifier: BSD-3-Clause
+# For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/BSD-3-Clause
+#
+import os
+import unittest
+import pandas as pd
+from omnixai.explainers.tabular import KNNCounterfactualExplainer
+from omnixai.tests.explainers.tasks import TabularClassification
+
+pd.set_option("display.max_columns", None)
+
+
+class TestKNNCE(unittest.TestCase):
+    def setUp(self):
+        base_folder = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..")
+        task = TabularClassification(base_folder).train_adult(num_training_samples=2000)
+        self.data = task.data
+        self.predict_function = lambda z: task.model.predict_proba(task.transform.transform(z))
+        self.test_instances = task.test_data.iloc(list(range(5))).remove_target_column()
+
+    def test_explain(self):
+        explainer = KNNCounterfactualExplainer(
+            training_data=self.data,
+            predict_function=self.predict_function,
+        )
+        explanations = explainer.explain(self.test_instances)
+        for explanation in explanations.get_explanations():
+            print("Query instance:")
+            print(explanation["query"])
+            print("Counterfactual examples:")
+            print(explanation["counterfactual"])
+            print("-----------------")
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/omnixai/tests/visualization/dashboard_tabular.py b/omnixai/tests/visualization/dashboard_tabular.py
index 36a31ca7..c36c0e42 100644
--- a/omnixai/tests/visualization/dashboard_tabular.py
+++ b/omnixai/tests/visualization/dashboard_tabular.py
@@ -98,7 +98,7 @@ def test(self):
         prediction_explanations = explainer.explain()
 
         explainers = TabularExplainer(
-            explainers=["lime", "shap", "mace", "pdp", "ale"],
+            explainers=["lime", "shap", "mace", "knn_ce", "pdp", "ale"],
             mode="classification",
             data=self.tabular_data,
             model=self.model,

From 21eed794904bcabb095008bd446195f8b8b0560f Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Wed, 16 Nov 2022 21:17:48 +0800
Subject: [PATCH 04/15] Add permutation feature importance

---
 omnixai/explainers/tabular/__init__.py        |   2 +
 .../tabular/agnostic/permutation.py           | 114 ++++++++++++++++++
 .../test_permutation_classification.py        |  25 ++++
 omnixai/tests/explainers/tasks.py             |   1 +
 4 files changed, 142 insertions(+)
 create mode 100644 omnixai/explainers/tabular/agnostic/permutation.py
 create mode 100644 omnixai/tests/explainers/permutation/test_permutation_classification.py

diff --git a/omnixai/explainers/tabular/__init__.py b/omnixai/explainers/tabular/__init__.py
index 31737c46..70057073 100644
--- a/omnixai/explainers/tabular/__init__.py
+++ b/omnixai/explainers/tabular/__init__.py
@@ -11,6 +11,7 @@
 from .agnostic.ale import ALE
 from .agnostic.sensitivity import SensitivityAnalysisTabular
 from .agnostic.L2X.l2x import L2XTabular
+from .agnostic.permutation import PermutationImportance
 from .counterfactual.mace.mace import MACEExplainer
 from .counterfactual.ce import CounterfactualExplainer
 from .counterfactual.knn import KNNCounterfactualExplainer
@@ -30,6 +31,7 @@
     "ALE",
     "SensitivityAnalysisTabular",
     "L2XTabular",
+    "PermutationImportance",
     "MACEExplainer",
     "CounterfactualExplainer",
     "KNNCounterfactualExplainer",
diff --git a/omnixai/explainers/tabular/agnostic/permutation.py b/omnixai/explainers/tabular/agnostic/permutation.py
new file mode 100644
index 00000000..44270689
--- /dev/null
+++ b/omnixai/explainers/tabular/agnostic/permutation.py
@@ -0,0 +1,114 @@
+#
+# Copyright (c) 2022 salesforce.com, inc.
+# All rights reserved.
+# SPDX-License-Identifier: BSD-3-Clause
+# For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/BSD-3-Clause
+#
+"""
+The permutation feature importance explanation for tabular data.
+"""
+import numpy as np
+import pandas as pd
+from typing import Callable
+from sklearn.metrics import log_loss
+from sklearn.inspection import permutation_importance
+
+from ..base import ExplainerBase, TabularExplainerMixin
+from ....data.tabular import Tabular
+from ....explanations.tabular.feature_importance import GlobalFeatureImportance
+
+
+class PermutationImportance(ExplainerBase, TabularExplainerMixin):
+    """
+    The permutation feature importance explanations for tabular data. The permutation feature
+    importance is defined to be the decrease in a model score when a single feature value
+    is randomly shuffled.
+    """
+
+    explanation_type = "global"
+    alias = ["permutation"]
+
+    def __init__(self, training_data: Tabular, predict_function, mode="classification", **kwargs):
+        """
+        :param training_data: The training dataset for training the machine learning model.
+        :param predict_function: The prediction function corresponding to the model to explain.
+            When the model is for classification, the outputs of the ``predict_function``
+            are the class probabilities. When the model is for regression, the outputs of
+            the ``predict_function`` are the estimated values.
+        :param mode: The task type, e.g., `classification` or `regression`.
+        """
+        super().__init__()
+        assert isinstance(training_data, Tabular), \
+            "training_data should be an instance of Tabular."
+        assert mode in ["classification", "regression"], \
+            "`mode` can only be `classification` or `regression`."
+
+        self.categorical_columns = training_data.categorical_columns
+        self.predict_function = predict_function
+        self.mode = mode
+
+    def _build_score_function(self, score_func=None):
+        if score_func is not None:
+            def _score(estimator, x, y):
+                z = self.predict_function(
+                    Tabular(x, categorical_columns=self.categorical_columns)
+                )
+                return score_func(z, y)
+        elif self.mode == "classification":
+            def _score(estimator, x, y):
+                z = self.predict_function(
+                    Tabular(x, categorical_columns=self.categorical_columns)
+                )
+                return log_loss(y, z)
+        else:
+            def _score(estimator, x, y):
+                z = self.predict_function(
+                    Tabular(x, categorical_columns=self.categorical_columns)
+                )
+                return np.mean((z - y) ** 2)
+        return _score
+
+    def explain(
+            self,
+            X,
+            y,
+            n_repeats: int = 30,
+            score_func: Callable = None
+    ) -> GlobalFeatureImportance:
+        """
+        Generate permutation feature importance scores.
+
+        :param X: Data on which permutation importance will be computed.
+        :param y: Targets or labels.
+        :param n_repeats: The number of times a feature is randomly shuffled.
+        :param score_func: The score/loss function measuring the difference between
+            predictions and ground-truth targets.
+        :return: The permutation feature importance explanations.
+        """
+        if isinstance(y, (list, tuple)):
+            y = np.array(y)
+        elif isinstance(y, pd.DataFrame):
+            y = y.values
+        elif isinstance(y, np.ndarray):
+            y = y
+        else:
+            raise ValueError(f"The type of `y` is {type(y)}, which is not supported."
+                             f"`y` should be a list, a numpy array or a pandas dataframe.")
+        if y.ndim > 1:
+            y = y.flatten()
+
+        assert X.shape[0] == len(y), \
+            "The numbers of samples in `X` and `y` are different."
+        X = X.remove_target_column()
+
+        class _Estimator:
+            def fit(self):
+                pass
+
+        results = permutation_importance(
+            estimator=_Estimator(),
+            X=X.to_pd(copy=False),
+            y=y,
+            scoring=self._build_score_function(score_func)
+        )
+        print(results)
diff --git a/omnixai/tests/explainers/permutation/test_permutation_classification.py b/omnixai/tests/explainers/permutation/test_permutation_classification.py
new file mode 100644
index 00000000..648a96cb
--- /dev/null
+++ b/omnixai/tests/explainers/permutation/test_permutation_classification.py
@@ -0,0 +1,25 @@
+#
+# Copyright (c) 2022 salesforce.com, inc.
+# All rights reserved.
+# SPDX-License-Identifier: BSD-3-Clause
+# For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/BSD-3-Clause
+#
+import os
+import unittest
+from omnixai.utils.misc import set_random_seed
+from omnixai.explainers.tabular import PermutationImportance
+from omnixai.tests.explainers.tasks import TabularClassification
+
+
+class TestPermutation(unittest.TestCase):
+    def test_1(self):
+        set_random_seed()
+        base_folder = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..")
+        task = TabularClassification(base_folder).train_adult(num_training_samples=2000)
+        predict_function = lambda z: task.model.predict_proba(task.transform.transform(z))
+        explainer = PermutationImportance(training_data=task.train_data, predict_function=predict_function)
+        explanations = explainer.explain(X=task.test_data, y=task.test_targets)
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/omnixai/tests/explainers/tasks.py b/omnixai/tests/explainers/tasks.py
index 6248c918..fa69aeb0 100644
--- a/omnixai/tests/explainers/tasks.py
+++ b/omnixai/tests/explainers/tasks.py
@@ -108,6 +108,7 @@ def train_adult(self, num_training_samples=None):
             data=tabular_data,
             train_data=transformer.invert(train),
             test_data=transformer.invert(test),
+            test_targets=labels_test
         )
 
     def train_iris(self):

From dc98ca68c688ee335ad0ac523e97730e0bd7b475 Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Wed, 16 Nov 2022 21:21:41 +0800
Subject: [PATCH 05/15] Add permutation feature importance

---
 omnixai/explainers/tabular/agnostic/permutation.py  | 13 +++++++++----
 .../permutation/test_permutation_classification.py  |  3 ++-
 2 files changed, 11 insertions(+), 5 deletions(-)

diff --git a/omnixai/explainers/tabular/agnostic/permutation.py b/omnixai/explainers/tabular/agnostic/permutation.py
index 44270689..6cd36cca 100644
--- a/omnixai/explainers/tabular/agnostic/permutation.py
+++ b/omnixai/explainers/tabular/agnostic/permutation.py
@@ -59,13 +59,13 @@ def _score(estimator, x, y):
                 z = self.predict_function(
                     Tabular(x, categorical_columns=self.categorical_columns)
                 )
-                return log_loss(y, z)
+                return -log_loss(y, z)
         else:
             def _score(estimator, x, y):
                 z = self.predict_function(
                     Tabular(x, categorical_columns=self.categorical_columns)
                 )
-                return np.mean((z - y) ** 2)
+                return -np.mean((z - y) ** 2)
         return _score
 
     def explain(
@@ -81,7 +81,7 @@ def explain(
         :param X: Data on which permutation importance will be computed.
         :param y: Targets or labels.
         :param n_repeats: The number of times a feature is randomly shuffled.
-        :param score_func: The score/loss function measuring the difference between
+        :param score_func: The score function measuring the difference between
             predictions and ground-truth targets.
         :return: The permutation feature importance explanations.
         """
@@ -105,10 +105,15 @@ class _Estimator:
             def fit(self):
                 pass
 
+        explanations = GlobalFeatureImportance()
         results = permutation_importance(
             estimator=_Estimator(),
             X=X.to_pd(copy=False),
             y=y,
             scoring=self._build_score_function(score_func)
         )
-        print(results)
+        explanations.add(
+            feature_names=list(X.columns),
+            importance_scores=results["importances_mean"]
+        )
+        return explanations
diff --git a/omnixai/tests/explainers/permutation/test_permutation_classification.py b/omnixai/tests/explainers/permutation/test_permutation_classification.py
index 648a96cb..70d0a793 100644
--- a/omnixai/tests/explainers/permutation/test_permutation_classification.py
+++ b/omnixai/tests/explainers/permutation/test_permutation_classification.py
@@ -12,13 +12,14 @@
 
 
 class TestPermutation(unittest.TestCase):
-    def test_1(self):
+    def test(self):
         set_random_seed()
         base_folder = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..")
         task = TabularClassification(base_folder).train_adult(num_training_samples=2000)
         predict_function = lambda z: task.model.predict_proba(task.transform.transform(z))
         explainer = PermutationImportance(training_data=task.train_data, predict_function=predict_function)
         explanations = explainer.explain(X=task.test_data, y=task.test_targets)
+        explanations.ipython_plot()
 
 
 if __name__ == "__main__":

From 56cf461e393ceb6a2167a91a7afbaf0a6f51ef41 Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Wed, 16 Nov 2022 21:24:46 +0800
Subject: [PATCH 06/15] Add permutation feature importance

---
 .../tabular/agnostic/permutation.py           |  2 +-
 .../test_permutation_classification.py        |  2 +-
 .../test_permutation_regression.py            | 26 +++++++++++++++++++
 3 files changed, 28 insertions(+), 2 deletions(-)
 create mode 100644 omnixai/tests/explainers/permutation/test_permutation_regression.py

diff --git a/omnixai/explainers/tabular/agnostic/permutation.py b/omnixai/explainers/tabular/agnostic/permutation.py
index 6cd36cca..eb824197 100644
--- a/omnixai/explainers/tabular/agnostic/permutation.py
+++ b/omnixai/explainers/tabular/agnostic/permutation.py
@@ -105,13 +105,13 @@ class _Estimator:
             def fit(self):
                 pass
 
-        explanations = GlobalFeatureImportance()
         results = permutation_importance(
             estimator=_Estimator(),
             X=X.to_pd(copy=False),
             y=y,
             scoring=self._build_score_function(score_func)
         )
+        explanations = GlobalFeatureImportance()
         explanations.add(
             feature_names=list(X.columns),
             importance_scores=results["importances_mean"]
diff --git a/omnixai/tests/explainers/permutation/test_permutation_classification.py b/omnixai/tests/explainers/permutation/test_permutation_classification.py
index 70d0a793..76db3828 100644
--- a/omnixai/tests/explainers/permutation/test_permutation_classification.py
+++ b/omnixai/tests/explainers/permutation/test_permutation_classification.py
@@ -19,7 +19,7 @@ def test(self):
         predict_function = lambda z: task.model.predict_proba(task.transform.transform(z))
         explainer = PermutationImportance(training_data=task.train_data, predict_function=predict_function)
         explanations = explainer.explain(X=task.test_data, y=task.test_targets)
-        explanations.ipython_plot()
+        explanations.plotly_plot()
 
 
 if __name__ == "__main__":
diff --git a/omnixai/tests/explainers/permutation/test_permutation_regression.py b/omnixai/tests/explainers/permutation/test_permutation_regression.py
new file mode 100644
index 00000000..dc815269
--- /dev/null
+++ b/omnixai/tests/explainers/permutation/test_permutation_regression.py
@@ -0,0 +1,26 @@
+#
+# Copyright (c) 2022 salesforce.com, inc.
+# All rights reserved.
+# SPDX-License-Identifier: BSD-3-Clause
+# For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/BSD-3-Clause
+#
+import unittest
+from omnixai.utils.misc import set_random_seed
+from omnixai.explainers.tabular import PermutationImportance
+from omnixai.tests.explainers.tasks import TabularRegression
+
+
+class TestPDPTabular(unittest.TestCase):
+    def test(self):
+        set_random_seed()
+        task = TabularRegression().train_boston()
+        predict_function = lambda z: task.model.predict(task.transform.transform(z))
+        explainer = PermutationImportance(
+            training_data=task.train_data, predict_function=predict_function, mode="regression"
+        )
+        explanations = explainer.explain(X=task.test_data, y=task.test_targets)
+        explanations.plotly_plot()
+
+
+if __name__ == "__main__":
+    unittest.main()

From 110ed3110e98cff730b325c0e4b8ee225c8b7775 Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Wed, 16 Nov 2022 21:26:28 +0800
Subject: [PATCH 07/15] Add permutation feature importance

---
 omnixai/explainers/tabular/agnostic/permutation.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/omnixai/explainers/tabular/agnostic/permutation.py b/omnixai/explainers/tabular/agnostic/permutation.py
index eb824197..4f7dfbb8 100644
--- a/omnixai/explainers/tabular/agnostic/permutation.py
+++ b/omnixai/explainers/tabular/agnostic/permutation.py
@@ -53,7 +53,7 @@ def _score(estimator, x, y):
                 z = self.predict_function(
                     Tabular(x, categorical_columns=self.categorical_columns)
                 )
-                return score_func(z, y)
+                return score_func(y, z)
         elif self.mode == "classification":
             def _score(estimator, x, y):
                 z = self.predict_function(
@@ -82,7 +82,7 @@ def explain(
         :param y: Targets or labels.
         :param n_repeats: The number of times a feature is randomly shuffled.
         :param score_func: The score function measuring the difference between
-            predictions and ground-truth targets.
+            ground-truth targets and predictions, e.g., -sklearn.metrics.log_loss(y_true, y_pred).
         :return: The permutation feature importance explanations.
         """
         if isinstance(y, (list, tuple)):

From fd91b3370c2a74b527fd70cd2b12e1cfbdff37df Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Wed, 16 Nov 2022 21:29:01 +0800
Subject: [PATCH 08/15] Add permutation feature importance

---
 omnixai/explainers/tabular/agnostic/permutation.py | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/omnixai/explainers/tabular/agnostic/permutation.py b/omnixai/explainers/tabular/agnostic/permutation.py
index 4f7dfbb8..b96c0e8a 100644
--- a/omnixai/explainers/tabular/agnostic/permutation.py
+++ b/omnixai/explainers/tabular/agnostic/permutation.py
@@ -9,7 +9,7 @@
 """
 import numpy as np
 import pandas as pd
-from typing import Callable
+from typing import Callable, Union
 from sklearn.metrics import log_loss
 from sklearn.inspection import permutation_importance
 
@@ -70,8 +70,8 @@ def _score(estimator, x, y):
 
     def explain(
             self,
-            X,
-            y,
+            X: Tabular,
+            y: Union[np.ndarray, pd.DataFrame],
             n_repeats: int = 30,
             score_func: Callable = None
     ) -> GlobalFeatureImportance:
@@ -85,6 +85,8 @@ def explain(
             ground-truth targets and predictions, e.g., -sklearn.metrics.log_loss(y_true, y_pred).
         :return: The permutation feature importance explanations.
         """
+        assert X is not None and y is not None, \
+            "The test data `X` and target `y` cannot be None."
         if isinstance(y, (list, tuple)):
             y = np.array(y)
         elif isinstance(y, pd.DataFrame):

From ccc6875b85ac7658a318812373de2cb6bec9a350 Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Wed, 16 Nov 2022 21:30:53 +0800
Subject: [PATCH 09/15] Add permutation feature importance

---
 omnixai/explainers/tabular/agnostic/permutation.py | 11 +----------
 1 file changed, 1 insertion(+), 10 deletions(-)

diff --git a/omnixai/explainers/tabular/agnostic/permutation.py b/omnixai/explainers/tabular/agnostic/permutation.py
index b96c0e8a..acf2019f 100644
--- a/omnixai/explainers/tabular/agnostic/permutation.py
+++ b/omnixai/explainers/tabular/agnostic/permutation.py
@@ -87,18 +87,9 @@ def explain(
         """
         assert X is not None and y is not None, \
             "The test data `X` and target `y` cannot be None."
-        if isinstance(y, (list, tuple)):
-            y = np.array(y)
-        elif isinstance(y, pd.DataFrame):
-            y = y.values
-        elif isinstance(y, np.ndarray):
-            y = y
-        else:
-            raise ValueError(f"The type of `y` is {type(y)}, which is not supported."
-                             f"`y` should be a list, a numpy array or a pandas dataframe.")
+        y = y.values if isinstance(y, pd.DataFrame) else np.array(y)
         if y.ndim > 1:
             y = y.flatten()
-
         assert X.shape[0] == len(y), \
             "The numbers of samples in `X` and `y` are different."
         X = X.remove_target_column()

From 0db40a394cd87be209382e41fb41c6016dd3ae58 Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Wed, 16 Nov 2022 21:33:16 +0800
Subject: [PATCH 10/15] Add permutation feature importance

---
 omnixai/explainers/tabular/agnostic/permutation.py | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/omnixai/explainers/tabular/agnostic/permutation.py b/omnixai/explainers/tabular/agnostic/permutation.py
index acf2019f..a629559a 100644
--- a/omnixai/explainers/tabular/agnostic/permutation.py
+++ b/omnixai/explainers/tabular/agnostic/permutation.py
@@ -18,6 +18,11 @@
 from ....explanations.tabular.feature_importance import GlobalFeatureImportance
 
 
+class _Estimator:
+    def fit(self):
+        pass
+
+
 class PermutationImportance(ExplainerBase, TabularExplainerMixin):
     """
     The permutation feature importance explanations for tabular data. The permutation feature
@@ -94,10 +99,6 @@ def explain(
             "The numbers of samples in `X` and `y` are different."
         X = X.remove_target_column()
 
-        class _Estimator:
-            def fit(self):
-                pass
-
         results = permutation_importance(
             estimator=_Estimator(),
             X=X.to_pd(copy=False),

From fddf217f0090210f25b7151f7f171c4aa7be2636 Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Thu, 17 Nov 2022 11:21:25 +0800
Subject: [PATCH 11/15] Add global Shap values

---
 omnixai/explainers/tabular/__init__.py        |   2 +
 omnixai/explainers/tabular/agnostic/shap.py   |  63 ++-------
 .../tabular/agnostic/shap_global.py           | 122 ++++++++++++++++++
 .../shap/test_shap_tabular_global.py          |   4 +-
 .../tests/visualization/dashboard_tabular.py  |   2 +-
 5 files changed, 141 insertions(+), 52 deletions(-)
 create mode 100644 omnixai/explainers/tabular/agnostic/shap_global.py

diff --git a/omnixai/explainers/tabular/__init__.py b/omnixai/explainers/tabular/__init__.py
index 70057073..4381ad65 100644
--- a/omnixai/explainers/tabular/__init__.py
+++ b/omnixai/explainers/tabular/__init__.py
@@ -12,6 +12,7 @@
 from .agnostic.sensitivity import SensitivityAnalysisTabular
 from .agnostic.L2X.l2x import L2XTabular
 from .agnostic.permutation import PermutationImportance
+from .agnostic.shap_global import GlobalShapTabular
 from .counterfactual.mace.mace import MACEExplainer
 from .counterfactual.ce import CounterfactualExplainer
 from .counterfactual.knn import KNNCounterfactualExplainer
@@ -32,6 +33,7 @@
     "SensitivityAnalysisTabular",
     "L2XTabular",
     "PermutationImportance",
+    "GlobalShapTabular",
     "MACEExplainer",
     "CounterfactualExplainer",
     "KNNCounterfactualExplainer",
diff --git a/omnixai/explainers/tabular/agnostic/shap.py b/omnixai/explainers/tabular/agnostic/shap.py
index e29abb36..535273f2 100644
--- a/omnixai/explainers/tabular/agnostic/shap.py
+++ b/omnixai/explainers/tabular/agnostic/shap.py
@@ -14,7 +14,7 @@
 from ..base import TabularExplainer
 from ....data.tabular import Tabular
 from ....explanations.tabular.feature_importance import \
-    FeatureImportance, GlobalFeatureImportance
+    FeatureImportance
 
 
 class ShapTabular(TabularExplainer):
@@ -23,7 +23,7 @@ class ShapTabular(TabularExplainer):
     If using this explainer, please cite the original work: https://github.com/slundberg/shap.
     """
 
-    explanation_type = "both"
+    explanation_type = "local"
     alias = ["shap"]
 
     def __init__(
@@ -58,34 +58,19 @@ def __init__(
             kwargs["nsamples"] = 100
         self.background_data = shap.sample(self.data, nsamples=kwargs["nsamples"])
 
-    def _explain_global(self, **kwargs) -> GlobalFeatureImportance:
-        if "nsamples" not in kwargs:
-            kwargs["nsamples"] = 100
-
-        instances = self.background_data
-        explanations = GlobalFeatureImportance()
-        explainer = shap.KernelExplainer(
-            self.predict_fn, self.background_data,
-            link="logit" if self.mode == "classification" else "identity", **kwargs
-        )
-        shap_values = explainer.shap_values(instances, **kwargs)
-
-        if self.mode == "classification":
-            values = 0
-            for v in shap_values:
-                values += np.abs(v)
-            values /= len(shap_values)
-            shap_values = values
-
-        importance_scores = np.mean(np.abs(shap_values), axis=0)
-        explanations.add(
-            feature_names=self.feature_columns,
-            importance_scores=importance_scores,
-            sort=True
-        )
-        return explanations
+    def explain(self, X, y=None, **kwargs) -> FeatureImportance:
+        """
+        Generates the local SHAP explanations for the input instances.
 
-    def _explain_local(self, X, y=None, **kwargs) -> FeatureImportance:
+        :param X: A batch of input instances. When ``X`` is `pd.DataFrame`
+            or `np.ndarray`, ``X`` will be converted into `Tabular` automatically.
+        :param y: A batch of labels to explain. For regression, ``y`` is ignored.
+            For classification, the top predicted label of each instance will be explained
+            when ``y = None``.
+        :param kwargs: Additional parameters for `shap.KernelExplainer.shap_values`,
+            e.g., ``nsamples`` -- the number of times to re-evaluate the model when explaining each prediction.
+        :return: The feature importance explanations.
+        """
         X = self._to_tabular(X).remove_target_column()
         explanations = FeatureImportance(self.mode)
         instances = self.transformer.transform(X)
@@ -165,26 +150,6 @@ def _predict(_x):
                 )
         return explanations
 
-    def explain(self, X=None, y=None, **kwargs):
-        """
-        Generates the local SHAP explanations for the input instances or global SHAP explanations.
-
-        :param X: A batch of input instances. When ``X`` is `pd.DataFrame`
-            or `np.ndarray`, ``X`` will be converted into `Tabular` automatically. If X is None,
-            it will compute global SHAP values, i.e., the mean of the absolute SHAP value for
-            each feature.
-        :param y: A batch of labels to explain. For regression, ``y`` is ignored.
-            For classification, the top predicted label of each instance will be explained
-            when ``y = None``.
-        :param kwargs: Additional parameters for `shap.KernelExplainer.shap_values`,
-            e.g., ``nsamples`` -- the number of times to re-evaluate the model when explaining each prediction.
-        :return: The feature-importance explanations.
-        """
-        if X is not None:
-            return self._explain_local(X=X, y=y, **kwargs)
-        else:
-            return self._explain_global(**kwargs)
-
     def save(
             self,
             directory: str,
diff --git a/omnixai/explainers/tabular/agnostic/shap_global.py b/omnixai/explainers/tabular/agnostic/shap_global.py
new file mode 100644
index 00000000..8ff9ab69
--- /dev/null
+++ b/omnixai/explainers/tabular/agnostic/shap_global.py
@@ -0,0 +1,122 @@
+#
+# Copyright (c) 2022 salesforce.com, inc.
+# All rights reserved.
+# SPDX-License-Identifier: BSD-3-Clause
+# For full license text, see the LICENSE file in the repo root or https://opensource.org/licenses/BSD-3-Clause
+#
+"""
+The SHAP explainer for global feature importance.
+"""
+import shap
+import numpy as np
+from typing import Callable, List
+
+from ..base import TabularExplainer
+from ....data.tabular import Tabular
+from ....explanations.tabular.feature_importance import GlobalFeatureImportance
+
+
+class GlobalShapTabular(TabularExplainer):
+    """
+    The SHAP explainer for global feature importance.
+    If using this explainer, please cite the original work: https://github.com/slundberg/shap.
+    """
+
+    explanation_type = "both"
+    alias = ["shap_global"]
+
+    def __init__(
+            self,
+            training_data: Tabular,
+            predict_function: Callable,
+            mode: str = "classification",
+            ignored_features: List = None,
+            **kwargs
+    ):
+        """
+        :param training_data: The data used to initialize a SHAP explainer. ``training_data``
+            can be the training dataset for training the machine learning model. If the training
+            dataset is large, please set parameter ``nsamples``, e.g., ``nsamples = 100``.
+        :param predict_function: The prediction function corresponding to the model to explain.
+            When the model is for classification, the outputs of the ``predict_function``
+            are the class probabilities. When the model is for regression, the outputs of
+            the ``predict_function`` are the estimated values.
+        :param mode: The task type, e.g., `classification` or `regression`.
+        :param ignored_features: The features ignored in computing feature importance scores.
+        :param kwargs: Additional parameters to initialize `shap.KernelExplainer`, e.g., ``nsamples``.
+            Please refer to the doc of `shap.KernelExplainer`.
+        """
+        super().__init__(training_data=training_data, predict_function=predict_function, mode=mode, **kwargs)
+        self.ignored_features = set(ignored_features) if ignored_features is not None else set()
+        if self.target_column is not None:
+            assert self.target_column not in self.ignored_features, \
+                f"The target column {self.target_column} cannot be in the ignored feature list."
+        self.valid_indices = [i for i, f in enumerate(self.feature_columns) if f not in self.ignored_features]
+
+        if "nsamples" not in kwargs:
+            kwargs["nsamples"] = 100
+        self.background_data = shap.sample(self.data, nsamples=kwargs["nsamples"])
+        self.sampled_data = shap.sample(self.data, nsamples=kwargs["nsamples"])
+
+    def _explain_global(self, X, **kwargs) -> GlobalFeatureImportance:
+        if "nsamples" not in kwargs:
+            kwargs["nsamples"] = 100
+        instances = self.sampled_data if X is None else \
+            self.transformer.transform(X.remove_target_column())
+
+        explanations = GlobalFeatureImportance()
+        explainer = shap.KernelExplainer(
+            self.predict_fn, self.background_data,
+            link="logit" if self.mode == "classification" else "identity", **kwargs
+        )
+        shap_values = explainer.shap_values(instances, **kwargs)
+
+        if self.mode == "classification":
+            values = 0
+            for v in shap_values:
+                values += np.abs(v)
+            values /= len(shap_values)
+            shap_values = values
+
+        importance_scores = np.mean(np.abs(shap_values), axis=0)
+        explanations.add(
+            feature_names=self.feature_columns,
+            importance_scores=importance_scores,
+            sort=True
+        )
+        return explanations
+
+    def explain(
+            self,
+            X: Tabular = None,
+            **kwargs
+    ):
+        """
+        Generates the global SHAP explanations.
+
+        :param X: The data will be used to compute global SHAP values, i.e., the mean of the absolute
+            SHAP value for each feature. If `X` is None, a set of training samples will be used.
+        :param kwargs: Additional parameters for `shap.KernelExplainer.shap_values`,
+            e.g., ``nsamples`` -- the number of times to re-evaluate the model when explaining each prediction.
+        :return: The global feature importance explanations.
+        """
+        return self._explain_global(X=X, **kwargs)
+
+    def save(
+            self,
+            directory: str,
+            filename: str = None,
+            **kwargs
+    ):
+        """
+        Saves the initialized explainer.
+
+        :param directory: The folder for the dumped explainer.
+        :param filename: The filename (the explainer class name if it is None).
+        """
+        super().save(
+            directory=directory,
+            filename=filename,
+            ignored_attributes=["data"],
+            **kwargs
+        )
diff --git a/omnixai/tests/explainers/shap/test_shap_tabular_global.py b/omnixai/tests/explainers/shap/test_shap_tabular_global.py
index 9b1d3938..caa3bd85 100644
--- a/omnixai/tests/explainers/shap/test_shap_tabular_global.py
+++ b/omnixai/tests/explainers/shap/test_shap_tabular_global.py
@@ -8,7 +8,7 @@
 import unittest
 import pprint
 from omnixai.utils.misc import set_random_seed
-from omnixai.explainers.tabular import ShapTabular
+from omnixai.explainers.tabular import GlobalShapTabular
 from omnixai.tests.explainers.tasks import TabularClassification
 
 
@@ -19,7 +19,7 @@ def test(self):
         predict_function = lambda z: task.model.predict_proba(task.transform.transform(z))
 
         set_random_seed()
-        explainer = ShapTabular(
+        explainer = GlobalShapTabular(
             training_data=task.train_data,
             predict_function=predict_function,
             ignored_features=None,
diff --git a/omnixai/tests/visualization/dashboard_tabular.py b/omnixai/tests/visualization/dashboard_tabular.py
index c36c0e42..7bf626fd 100644
--- a/omnixai/tests/visualization/dashboard_tabular.py
+++ b/omnixai/tests/visualization/dashboard_tabular.py
@@ -98,7 +98,7 @@ def test(self):
         prediction_explanations = explainer.explain()
 
         explainers = TabularExplainer(
-            explainers=["lime", "shap", "mace", "knn_ce", "pdp", "ale"],
+            explainers=["lime", "shap", "mace", "knn_ce", "pdp", "ale", "shap_global"],
             mode="classification",
             data=self.tabular_data,
             model=self.model,

From b35dacfee9d2dc76018cd26e0ed88819ce8590c5 Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Thu, 17 Nov 2022 11:28:35 +0800
Subject: [PATCH 12/15] Add global Shap values

---
 omnixai/explainers/tabular/agnostic/shap_global.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/omnixai/explainers/tabular/agnostic/shap_global.py b/omnixai/explainers/tabular/agnostic/shap_global.py
index 8ff9ab69..2ee94b64 100644
--- a/omnixai/explainers/tabular/agnostic/shap_global.py
+++ b/omnixai/explainers/tabular/agnostic/shap_global.py
@@ -22,7 +22,7 @@ class GlobalShapTabular(TabularExplainer):
     If using this explainer, please cite the original work: https://github.com/slundberg/shap.
     """
 
-    explanation_type = "both"
+    explanation_type = "global"
     alias = ["shap_global"]
 
     def __init__(

From ce94c1556c387f316116fb87f8fc45176affc57c Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Thu, 17 Nov 2022 15:58:51 +0800
Subject: [PATCH 13/15] Revise dashboard figures

---
 omnixai/explanations/base.py | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/omnixai/explanations/base.py b/omnixai/explanations/base.py
index e1748232..c8f4f36f 100644
--- a/omnixai/explanations/base.py
+++ b/omnixai/explanations/base.py
@@ -137,7 +137,16 @@ def to_html_div(self, id=None):
         elif isinstance(self.component, dash_table.DataTable):
             return html.Div([self.component], id=id)
         elif isinstance(self.component, plotly.graph_objs.Figure):
-            return html.Div([dcc.Graph(figure=self.component, id=id)], id=f"div_{id}")
+            height = self.component.layout.height
+            if height is None or height <= 450:
+                return html.Div(
+                    [dcc.Graph(figure=self.component, id=id)],
+                    id=f"div_{id}")
+            else:
+                return html.Div(
+                    [dcc.Graph(figure=self.component, id=id)],
+                    id=f"div_{id}",
+                    style={"overflowY": "scroll", "height": 450})
         else:
             raise ValueError(f"The type of `component` ({type(self.component)}) "
                              f"" f"is not supported by DashFigure.")

From f0240794e53674f6c90374ef1c91f5aa9104ba03 Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Thu, 17 Nov 2022 16:16:07 +0800
Subject: [PATCH 14/15] Revise SHAP parameters

---
 omnixai/explainers/tabular/agnostic/shap.py | 20 ++++++++------------
 1 file changed, 8 insertions(+), 12 deletions(-)

diff --git a/omnixai/explainers/tabular/agnostic/shap.py b/omnixai/explainers/tabular/agnostic/shap.py
index 535273f2..e283ed8f 100644
--- a/omnixai/explainers/tabular/agnostic/shap.py
+++ b/omnixai/explainers/tabular/agnostic/shap.py
@@ -22,7 +22,6 @@ class ShapTabular(TabularExplainer):
     The SHAP explainer for tabular data.
     If using this explainer, please cite the original work: https://github.com/slundberg/shap.
     """
-
     explanation_type = "local"
     alias = ["shap"]
 
@@ -48,15 +47,18 @@ def __init__(
             Please refer to the doc of `shap.KernelExplainer`.
         """
         super().__init__(training_data=training_data, predict_function=predict_function, mode=mode, **kwargs)
+        self.link = kwargs.get("link", None)
+        if self.link is None:
+            self.link = "logit" if self.mode == "classification" else "identity"
+
         self.ignored_features = set(ignored_features) if ignored_features is not None else set()
         if self.target_column is not None:
             assert self.target_column not in self.ignored_features, \
                 f"The target column {self.target_column} cannot be in the ignored feature list."
         self.valid_indices = [i for i, f in enumerate(self.feature_columns) if f not in self.ignored_features]
 
-        if "nsamples" not in kwargs:
-            kwargs["nsamples"] = 100
-        self.background_data = shap.sample(self.data, nsamples=kwargs["nsamples"])
+        self.background_data = shap.sample(self.data, nsamples=kwargs.get("nsamples", 100))
+        self.explainer = shap.KernelExplainer(self.predict_fn, self.background_data, link=self.link, **kwargs)
 
     def explain(self, X, y=None, **kwargs) -> FeatureImportance:
         """
@@ -91,12 +93,7 @@ def explain(self, X, y=None, **kwargs) -> FeatureImportance:
             y = None
 
         if len(self.ignored_features) == 0:
-            explainer = shap.KernelExplainer(
-                self.predict_fn, self.background_data,
-                link="logit" if self.mode == "classification" else "identity", **kwargs
-            )
-            shap_values = explainer.shap_values(instances, **kwargs)
-
+            shap_values = self.explainer.shap_values(instances, **kwargs)
             for i, instance in enumerate(instances):
                 df = X.iloc(i).to_pd()
                 feature_values = \
@@ -124,10 +121,9 @@ def _predict(_x):
 
                 predict_function = _predict
                 test_x = instance[self.valid_indices]
-
                 explainer = shap.KernelExplainer(
                     predict_function, self.background_data[:, self.valid_indices],
-                    link="logit" if self.mode == "classification" else "identity", **kwargs
+                    link=self.link, **kwargs
                 )
                 shap_values = explainer.shap_values(np.expand_dims(test_x, axis=0), **kwargs)
 

From d3ebd0e3a4d0ddd79381aa84b8ad3119838289fd Mon Sep 17 00:00:00 2001
From: ywz <yangwenzhuo08@gmail.com>
Date: Thu, 17 Nov 2022 16:39:55 +0800
Subject: [PATCH 15/15] Update docs

---
 README.md                                    |  1 +
 docs/omnixai.explainers.tabular.agnostic.rst | 16 ++++++++++++++++
 2 files changed, 17 insertions(+)

diff --git a/README.md b/README.md
index 01beadcb..356b5856 100644
--- a/README.md
+++ b/README.md
@@ -63,6 +63,7 @@ We will continue improving this library to make it more comprehensive in the fut
 | Partial dependence plots  |   Black box   |      Global      |     |    ✅    |       |      |      |
 | Accumulated local effects |   Black box   |      Global      |     |    ✅    |       |      |      |
 |   Sensitivity analysis    |   Black box   |      Global      |     |    ✅    |       |      |      |
+|  Permutation explanation  |   Black box   |      Global      |     |    ✅    |       |      |      |
 |   Feature visualization   |  Torch or TF  |      Global      |     |         |   ✅   |      |      |
 |       Feature maps        |  Torch or TF  |      Local       |     |         |   ✅   |      |      |
 |           LIME            |   Black box   |      Local       |     |    ✅    |   ✅   | ✅   |      |
diff --git a/docs/omnixai.explainers.tabular.agnostic.rst b/docs/omnixai.explainers.tabular.agnostic.rst
index 3597ced6..ee8dbfca 100644
--- a/docs/omnixai.explainers.tabular.agnostic.rst
+++ b/docs/omnixai.explainers.tabular.agnostic.rst
@@ -60,3 +60,19 @@ omnixai.explainers.tabular.agnostic.L2X.l2x module
    :members:
    :undoc-members:
    :show-inheritance:
+
+omnixai.explainers.tabular.agnostic.permutation module
+------------------------------------------------------
+
+.. automodule:: omnixai.explainers.tabular.agnostic.permutation
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+omnixai.explainers.tabular.agnostic.shap_global module
+------------------------------------------------------
+
+.. automodule:: omnixai.explainers.tabular.agnostic.shap_global
+   :members:
+   :undoc-members:
+   :show-inheritance: