Closed SyfSchydea closed 4 years ago
When a cluster is empty, it should keep the same position as its previous iteration.
Currently the cluster is re-assigned to a random point, but this seems to result in infinite loops on some data-sets.
This method would keep k-means in-line with scipy's k-means implementation (when it doesn't simply raise an error for it)
Done
When a cluster is empty, it should keep the same position as its previous iteration.
Currently the cluster is re-assigned to a random point, but this seems to result in infinite loops on some data-sets.
This method would keep k-means in-line with scipy's k-means implementation (when it doesn't simply raise an error for it)