replace kmeans() with kmeans2()

This might be a solution for issue #10

Author: slowkow

CHANGELOG.md

@@ -1,3 +1,10 @@
+# 0.0.6 - 2022-02-02
+
+- Replace `scipy.cluster.vq.kmeans` with `scipy.cluster.vq.kmeans2` to address
+  issue #10 where we learned that kmeans does not always return k centroids,
+  but kmeans2 does return k centroids. Thanks to @onionpork and @DennisPost10
+  for reporting this.
+
 # 0.0.5 - 2020-08-11
 
 - Expose `max_iter_harmony` as a new top-level argument, in addition to the

harmonypy/harmony.py

@@ -17,7 +17,8 @@
 
 import pandas as pd
 import numpy as np
-from scipy.cluster.vq import kmeans
+# kmeans does not always return k centroids, but kmeans2 does
+from scipy.cluster.vq import kmeans2
 import logging
 
 # create logger
@@ -185,8 +186,8 @@ def allocate_buffers(self):
 
     def init_cluster(self):
         # Start with cluster centroids
-        km = kmeans(self.Z_cos.T, self.K, iter=10)
-        self.Y = km[0].T
+        km_centroids, km_labels = kmeans2(self.Z_cos.T, self.K, minit='++')
+        self.Y = km_centroids.T
         # (1) Normalize
         self.Y = self.Y / np.linalg.norm(self.Y, ord=2, axis=0)
         # (2) Assign cluster probabilities

setup.py

@@ -5,7 +5,7 @@
 
 setuptools.setup(
     name                          = "harmonypy",
-    version                       = "0.0.5",
+    version                       = "0.0.6",
     author                        = "Kamil Slowikowski",
     author_email                  = "[email protected]",
     description                   = "A data integration algorithm.",