KMedoids supports array-like init method (#137)

* KMedoids supports array-like init method

* make input X as optional

* remove unnecessary data type checking

* fix formatting according to flake8 suggestions

* fix format

* add KMedoids array-like init test case for max_iter=0

* add test cases for n_clusters > number of centroids when using array-like init method

* lint formatting

sklearn_extra/cluster/_k_medoids.py

@@ -64,7 +64,8 @@ class KMedoids(BaseEstimator, ClusterMixin, TransformerMixin):
     method : {'alternate', 'pam'}, default: 'alternate'
         Which algorithm to use. 'alternate' is faster while 'pam' is more accurate.
 
-    init : {'random', 'heuristic', 'k-medoids++', 'build'}, optional, default: 'heuristic'
+    init : {'random', 'heuristic', 'k-medoids++', 'build'}, or array-like of shape
+        (n_clusters, n_features), optional, default: 'heuristic'
         Specify medoid initialization method. 'random' selects n_clusters
         elements from the dataset. 'heuristic' picks the n_clusters points
         with the smallest sum distance to every other point. 'k-medoids++'
@@ -74,6 +75,8 @@ class KMedoids(BaseEstimator, ClusterMixin, TransformerMixin):
         algorithm. Often 'build' is more efficient but slower than other
         initializations on big datasets and it is also very non-robust,
         if there are outliers in the dataset, use another initialization.
+        If an array is passed, it should be of shape (n_clusters, n_features)
+        and gives the initial centers.
 
         .. _k-means++: https://theory.stanford.edu/~sergei/papers/kMeansPP-soda.pdf
 
@@ -181,13 +184,29 @@ def _check_init_args(self):
 
         # Check init
         init_methods = ["random", "heuristic", "k-medoids++", "build"]
-        if self.init not in init_methods:
+        if not (
+            hasattr(self.init, "__array__")
+            or (isinstance(self.init, str) and self.init in init_methods)
+        ):
             raise ValueError(
                 "init needs to be one of "
                 + "the following: "
-                + "%s" % init_methods
+                + "%s" % (init_methods + ["array-like"])
             )
 
+        # Check n_clusters
+        if (
+            hasattr(self.init, "__array__")
+            and self.n_clusters != self.init.shape[0]
+        ):
+            warnings.warn(
+                "n_clusters should be equal to size of array-like if init "
+                "is array-like setting n_clusters to {}.".format(
+                    self.init.shape[0]
+                )
+            )
+            self.n_clusters = self.init.shape[0]
+
     def fit(self, X, y=None):
         """Fit K-Medoids to the provided data.
 
@@ -219,7 +238,7 @@ def fit(self, X, y=None):
         D = pairwise_distances(X, metric=self.metric)
 
         medoid_idxs = self._initialize_medoids(
-            D, self.n_clusters, random_state_
+            D, self.n_clusters, random_state_, X
         )
         labels = None
 
@@ -407,10 +426,14 @@ def predict(self, X):
 
             return pd_argmin
 
-    def _initialize_medoids(self, D, n_clusters, random_state_):
+    def _initialize_medoids(self, D, n_clusters, random_state_, X=None):
         """Select initial mediods when beginning clustering."""
 
-        if self.init == "random":  # Random initialization
+        if hasattr(self.init, "__array__"):  # Pre assign cluster
+            medoids = np.hstack(
+                [np.where((X == c).all(axis=1)) for c in self.init]
+            ).ravel()
+        elif self.init == "random":  # Random initialization
             # Pick random k medoids as the initial ones.
             medoids = random_state_.choice(len(D), n_clusters, replace=False)
         elif self.init == "k-medoids++":

sklearn_extra/cluster/tests/test_k_medoids.py

@@ -423,7 +423,33 @@ def test_medoids_indices():
 
     model = KMedoids(n_clusters=3, init="build", random_state=rng)
 
+    centroids = np.array([X_iris[0], X_iris[50]])
+    array_like_model = KMedoids(
+        n_clusters=len(centroids), init=centroids, max_iter=0
+    )
+
     model.fit(X_iris)
     clara.fit(X_iris)
+    array_like_model.fit(X_iris)
     assert_array_equal(X_iris[model.medoid_indices_], model.cluster_centers_)
     assert_array_equal(X_iris[clara.medoid_indices_], clara.cluster_centers_)
+    assert_array_equal(centroids, array_like_model.cluster_centers_)
+
+
+def test_array_like_init():
+    centroids = np.array([X_cc[0], X_cc[50]])
+
+    expected = np.hstack([np.zeros(50), np.ones(50)])
+    km = KMedoids(n_clusters=len(centroids), init=centroids)
+    km.fit(X_cc)
+    # # This test use data that are not perfectly separable so the
+    # # accuracy is not 1. Accuracy around 0.85
+    assert (np.mean(km.labels_ == expected) > 0.8) or (
+        1 - np.mean(km.labels_ == expected) > 0.8
+    )
+
+    # Override n_clusters if array-like init method is used
+    km = KMedoids(n_clusters=len(centroids) + 2, init=centroids)
+    km.fit(X_cc)
+
+    assert len(km.cluster_centers_) == len(centroids)