FIX BisectingKMeans crashes randomly (#25563)

jeremiedbb · jeremiedbb · commit ed37222729c1 · 2023-03-08T17:21:14.000+01:00
Fixes #25505
diff --git a/doc/whats_new/v1.2.rst b/doc/whats_new/v1.2.rst
@@ -27,6 +27,13 @@ Changes impacting all modules
 Changelog
 ---------
 
+:mod:`sklearn.cluster`
+......................
+
+- |Fix| Fixed a bug in :class:`cluster.BisectingKMeans`, preventing `fit` to randomly
+  fail due to a permutation of the labels when running multiple inits.
+  :pr:`25563` by :user:`Jérémie du Boisberranger <jeremiedbb>`.
+
 :mod:`sklearn.isotonic`
 .......................
 
diff --git a/sklearn/cluster/_bisect_k_means.py b/sklearn/cluster/_bisect_k_means.py
@@ -337,7 +337,9 @@ def _bisect(self, X, x_squared_norms, sample_weight, cluster_to_bisect):
                 X, best_centers, best_labels, sample_weight
             )
         else:  # bisecting_strategy == "largest_cluster"
-            scores = np.bincount(best_labels)
+            # Using minlength to make sure that we have the counts for both labels even
+            # if all samples are labelled 0.
+            scores = np.bincount(best_labels, minlength=2)
 
         cluster_to_bisect.split(best_labels, best_centers, scores)
 

Original file line number	Diff line number	Diff line change
`@@ -337,7 +337,9 @@ def _bisect(self, X, x_squared_norms, sample_weight, cluster_to_bisect):`
`337`	`337`	`X, best_centers, best_labels, sample_weight`
`338`	`338`	`)`
`339`	`339`	`else: # bisecting_strategy == "largest_cluster"`
`340`		`- scores = np.bincount(best_labels)`
	`340`	`+ # Using minlength to make sure that we have the counts for both labels even`
	`341`	`+ # if all samples are labelled 0.`
	`342`	`+ scores = np.bincount(best_labels, minlength=2)`
`341`	`343`
`342`	`344`	`cluster_to_bisect.split(best_labels, best_centers, scores)`
`343`	`345`