add centroid metric, fix calculating distance from hyperplane for mul…

…ticlass classification

fsvm/_fuzzy_svc.py

@@ -229,7 +229,7 @@ class FuzzySVC(ClassifierMixin, BaseEstimator):
            0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
            1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1, 1, 2, 1, 1, 1, 1,
            1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 1, 2, 2, 2,
-           2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+           2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 2, 2, 2, 2, 2,
            2, 2, 2, 2, 2, 2, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2])
     """
 
@@ -238,6 +238,7 @@ class FuzzySVC(ClassifierMixin, BaseEstimator):
             StrOptions({"centroid", "hyperplane"}),
             callable,
         ],
+        "centroid_metric": [StrOptions({"euclidean", "manhattan"})],
         "membership_decay": [StrOptions({"exponential", "linear"}), callable],
         "beta": [Interval(Real, 0.0, 1.0, closed="both")],
         "balanced": ["boolean"],
@@ -251,8 +252,9 @@ def __init__(
         self,
         *,
         distance_metric="centroid",
+        centroid_metric="euclidean",
         membership_decay="exponential",
-        beta=0.5,
+        beta=0.1,
         balanced=True,
         C=1.0,
         kernel="rbf",
@@ -270,6 +272,7 @@ def __init__(
         random_state=None,
     ):
         self.distance_metric = distance_metric
+        self.centroid_metric = centroid_metric
         self.membership_decay = membership_decay
         self.beta = beta
         self.balanced = balanced
@@ -349,8 +352,21 @@ def fit(self, X, y):
             centroids = _NearestCentroid().fit(X, y).centroids_
             self.distance_ = np.linalg.norm(X - centroids[y_], axis=1)
         elif self.distance_metric == "hyperplane":
-            svc = _SVC(**svc_args).fit(X, y)
-            self.distance_ = np.abs(svc.decision_function(X))
+            hyperplane_svc_args = {**svc_args, "decision_function_shape": "ovr"}
+            svc = _SVC(**hyperplane_svc_args).fit(X, y)
+            decision_function_output = svc.decision_function(X)
+
+            # For multiclass, extract the distances corresponding
+            # to the true class labels
+            if decision_function_output.ndim > 1:
+                y_indices = np.array(
+                    [svc.classes_.tolist().index(class_label) for class_label in y]
+                )
+                decision_function_output = decision_function_output[
+                    np.arange(len(y)), y_indices
+                ]
+
+            self.distance_ = np.abs(decision_function_output)
         elif callable(self.distance_metric):
             self.distance_ = self.distance_metric(X)
 
@@ -416,7 +432,7 @@ def __calculate_membership_degree(self):
             membership = 2 / (1 + np.exp(self.beta * self.distance_))
         elif self.membership_decay == "linear":
             max_distance = np.amax(self.distance_)
-            delta = 1e-6
+            delta = 1e-9
             membership = 1 - (self.distance_ / (max_distance + delta))
         elif callable(self.membership_decay):
             membership = self.membership_decay(self.distance_)

fsvm/tests/test_fuzzy_svc.py

@@ -17,7 +17,7 @@ def test_fuzzy_svc(data):
     clf = FuzzySVC()
     assert clf.distance_metric == "centroid"
     assert clf.membership_decay == "exponential"
-    assert clf.beta == 0.5
+    assert clf.beta == 0.1
     assert clf.balanced is True
 
     clf.fit(X, y)