formatting

.flake8

@@ -1,2 +1,3 @@
 [flake8]
 max-line-length = 127
+per-file-ignores = setup.py:F821

focus/__init__.py → cfepy/__init__.py

@@ -1,4 +1,4 @@
-from .core import Focus
+from cfepy.focus.core import Focus
 
 # Let users know if they're missing any of our hard dependencies
 _hard_dependencies = ("tensorflow", "numpy")

focus/core.py → cfepy/focus/core.py

@@ -16,7 +16,7 @@
 
 
 class Focus:
-    r"""
+    """
     FOCUS Lucic, et al. 2022 computes
     Counterfactual Explanations (CFE) using the gradient descent
     method for predictions of the tree-based models.
@@ -70,9 +70,9 @@ class Focus:
     - Initialize FOCUS with default parameters
     - Generate counterfactual explanations
 
-    focus-cfe = Focus()
+    cfepy = Focus()
 
-    cfe_features = focus-cfe.generate(model, X)
+    cfe_features = cfepy.generate(model, X)
     """
 
     def __init__(
@@ -100,7 +100,7 @@ def __init__(
         self.verbose = verbose
 
     def generate(self, model, X, x_train=None):
-        r"""
+        """
         Generate counterfactual explanations for the
         predictions from a tree-based model.
 
@@ -214,7 +214,7 @@ def generate(self, model, X, x_train=None):
 
     @staticmethod
     def prepare_features_by_perturb_direction(model, X: np.ndarray, direction: str):
-        r"""
+        """
         Prepares the input data `X` based on the perturbation direction.
 
         Args:
@@ -267,7 +267,7 @@ def compute_gradient(
         temperature,
         optimizer,
     ):
-        r"""
+        """
         Computes the gradient of the loss function with respect to the variables to optimize.
 
         Returns:
@@ -302,7 +302,7 @@ def compute_gradient(
 
     @staticmethod
     def parse_class_tree(tree, X, sigma: float) -> list:
-        r"""
+        """
         Compute impurity of each leaf node in a decision tree and approximate it using sigmoid function.
 
         Args:
@@ -341,7 +341,6 @@ def parse_class_tree(tree, X, sigma: float) -> list:
         for i in range(n_nodes):
             cur_node = nodes[i]
             if children_left[i] != children_right[i]:
-
                 if cur_node is None:
                     cur_node = 1.0
 
@@ -362,7 +361,7 @@ def parse_class_tree(tree, X, sigma: float) -> list:
 
     @staticmethod
     def get_prob_classification_tree(tree, X, sigma: float) -> tf.Tensor:
-        r"""
+        """
         get_prob_classification_tree - computes the probability of each sample's classification in a decision tree
 
         Args:
@@ -420,7 +419,7 @@ def get_prob_classification_tree(tree, X, sigma: float) -> tf.Tensor:
     def get_prob_classification_forest(
         model, X: tf.Tensor, sigma: float, temperature: float
     ) -> tf.Tensor:
-        r"""
+        """
         Calculate the softmax probabilities for classification for a random forest or AdaBoost model.
 
         Args:
@@ -472,7 +471,7 @@ def filter_hinge_loss(
         temperature,
         model,
     ) -> tf.Tensor:
-        r"""
+        """
         Calculates the filtered probabilities of each data point for the given model.
 
         Args:

focus/utils.py → cfepy/focus/utils.py

@@ -2,7 +2,7 @@
 import tensorflow as tf
 
 
-def safe_euclidean(matrix_diff, epsilon=10.0 ** -10) -> tf.Tensor:
+def safe_euclidean(matrix_diff, epsilon=10.0**-10) -> tf.Tensor:
     """
     Calculates the Euclidean distance between two matrices with a small epsilon added to prevent singularities.
 
@@ -13,10 +13,10 @@ def safe_euclidean(matrix_diff, epsilon=10.0 ** -10) -> tf.Tensor:
     Returns:
     tf.Tensor: A tensor representing the Euclidean distance between the two matrices
     """
-    return (tf.reduce_sum(matrix_diff ** 2, axis=-1) + epsilon) ** 0.5
+    return (tf.reduce_sum(matrix_diff**2, axis=-1) + epsilon) ** 0.5
 
 
-def safe_cosine(feat_input, perturbed, epsilon=10.0 ** -10) -> tf.Tensor:
+def safe_cosine(feat_input, perturbed, epsilon=10.0**-10) -> tf.Tensor:
     """
     Calculates cosine distance between two input arrays `feat_input` and `perturbed`
     while ensuring numerical stability with `epsilon`.
@@ -42,7 +42,7 @@ def safe_cosine(feat_input, perturbed, epsilon=10.0 ** -10) -> tf.Tensor:
     return dist
 
 
-def safe_l1(matrix_diff, epsilon=10.0 ** -10) -> tf.Tensor:
+def safe_l1(matrix_diff, epsilon=10.0**-10) -> tf.Tensor:
     """
     Calculates the L1 (Manhattan) distance between two tensors with a small epsilon value
     added to prevent division by zero.
@@ -76,7 +76,7 @@ def tf_cov(x_train) -> tf.Tensor:
     return cov_xx
 
 
-def safe_mahal(matrix_diff, x_train, epsilon=10.0 ** -10) -> tf.Tensor:
+def safe_mahal(matrix_diff, x_train, epsilon=10.0**-10) -> tf.Tensor:
     """
     Calculates Mahalanobis distance using TensorFlow
 

focus/tests/test_focus.py → cfepy/tests/test_focus.py

@@ -4,7 +4,7 @@
 from sklearn.ensemble import AdaBoostClassifier, RandomForestClassifier
 from sklearn.tree import DecisionTreeClassifier
 
-from focus import Focus
+from cfepy import Focus
 
 x_train, y_train = make_classification(
     n_samples=200, n_features=10, n_classes=2, random_state=42

focus/tests/test_utils.py → cfepy/tests/test_utils.py

@@ -4,7 +4,7 @@
 import pytest
 import tensorflow as tf
 
-from focus.utils import (
+from cfepy import (
     calculate_distance,
     safe_cosine,
     safe_euclidean,
@@ -13,7 +13,7 @@
     tf_cov,
 )
 
-epsilon = 10.0 ** -10
+epsilon = 10.0**-10
 random.seed(42)
 
 covariance_test_data = [
@@ -88,7 +88,7 @@ def test_tf_cov(feat_input_cov, expected_output_cov):
 
 @pytest.mark.parametrize("feat_input, feat_input2", distance_test_data)
 def test_safe_euclidean(feat_input, feat_input2):
-    expected = (np.sum(feat_input2 ** 2, axis=-1) + epsilon) ** 0.5
+    expected = (np.sum(feat_input2**2, axis=-1) + epsilon) ** 0.5
     assert safe_euclidean(feat_input).numpy() == pytest.approx(expected)
 
 

focus/version.py → cfepy/version.py

@@ -1,5 +1,5 @@
 """
-``focus-cfe`` is a python package for
+``cfepy`` is a python package for
 generating counterfactual explanations for a tree-based model
 """
 # PEP0440 compatible formatted version, see:
@@ -18,4 +18,4 @@
 # Dev branch marker is: 'X.Y.dev' or 'X.Y.devN' where N is an integer.
 # 'X.Y.dev0' is the canonical version of 'X.Y.dev'
 #
-__version__ = "0.0.dev3"  # pragma: no cover
+__version__ = "0.0.dev4"  # pragma: no cover

docs/conf.py

@@ -18,7 +18,7 @@
 
 # -- Project information -----------------------------------------------------
 
-project = "focus-cfe"
+project = "cfepy"
 copyright = "2023, Kyosuke Morita"
 author = "Kyosuke Morita"
 

environment.yml

@@ -1,4 +1,4 @@
-name: focus-cfe
+name: cfepy
 dependencies:
   - tensorflow
   - numpy

examples/focus_example.py

@@ -8,7 +8,7 @@
     train_decision_tree_model,
 )
 
-from focus import Focus
+from cfepy import Focus
 
 
 def run_example():

examples/hyperparameter_tuning_example.py

@@ -6,7 +6,7 @@
     train_decision_tree_model,
 )
 
-from focus import Focus
+from cfepy import Focus
 
 
 def objective(trial):
@@ -60,7 +60,6 @@ def objective(trial):
 
 
 if __name__ == "__main__":
-
     study = optuna.create_study(direction="minimize")
     study.optimize(objective, n_trials=100)
 

examples/utils.py

@@ -84,14 +84,14 @@ def prepare_plot_df(model, X, X_focus):
     Args:
         model (object): A trained machine learning model capable of making predictions.
         X (array-like): The input data for which predictions are made.
-        X_focus (array-like): Additional input data used for focus-cfe predictions.
+        X_focus (array-like): Additional input data used for cfepy predictions.
 
     Returns:
         tuple: A tuple containing two pandas DataFrames.
             - The first DataFrame contains the PCA-transformed features of `X`
                 and the corresponding predictions.
             - The second DataFrame contains the PCA-transformed features of `X_focus`
-                and the corresponding focus-cfe predictions.
+                and the corresponding cfepy predictions.
     """
     pca = PCA(n_components=2)
 

setup.py

@@ -10,7 +10,7 @@
 
 
 # get __version__ from _version.py
-ver_file = path.join("focus", "version.py")
+ver_file = path.join("cfepy", "version.py")
 with open(ver_file) as f:
     exec(f.read())
 
@@ -28,7 +28,7 @@ def readme():
     requirements = f.read().splitlines()
 
 setup(
-    name="focus-cfe",
+    name="cfepy",
     version=__version__,
     author="Kyosuke Morita",
     author_email="[email protected]",