Scikit-learn dont implement the arccos distance, use cosine distance to pairwise distance

Author: Bruno Alano

hdbscan/dist_metrics.pyx

@@ -1052,6 +1052,10 @@ cdef class CosineDistance(DistanceMetric):
             norm2 += x2[j] * x2[j]
         return 1.0 - d / sqrt(norm1 * norm2)
 
+#------------------------------------------------------------
+# Cosine Distance
+#  D(x, y) = arccos(dot(x, y) / (|x| * |y|)) / PI
+
 cdef class ArccosDistance(DistanceMetric):
     cdef inline DTYPE_t dist(self, DTYPE_t* x1, DTYPE_t* x2, ITYPE_t size) nogil except -1:
         cdef DTYPE_t d = 0, norm1 = 0, norm2 = 0

hdbscan/hdbscan_.py

@@ -68,6 +68,8 @@ def _hdbscan_generic(X, min_samples=5, alpha=1.0,
             raise ValueError('Minkowski metric with negative p value is not defined!')
 
         distance_matrix = pairwise_distances(X, metric=metric, p=p)
+    elif metric == 'arccos':
+        distance_matrix = pairwise_distances(X, metric='cosine', **kwargs)
     else:
         distance_matrix = pairwise_distances(X, metric=metric, **kwargs)