k-means should support better handling of empty clusters #595

rcurtin opened this Issue Mar 26, 2016 · 1 comment


None yet

1 participant

rcurtin commented Mar 26, 2016

We should update the KMeans class so that the EmptyClusterPolicy and only the EmptyClusterPolicy controls what is done with empty clusters. Two options I plan to implement are:

  • "kill" the cluster by setting it to all DBL_MAX, so it's never used again
  • leave the cluster alone, by simply leaving the centroid in place; this is Erich's (@kno10) suggestion

This comes out of the discussion for #592.

@rcurtin rcurtin self-assigned this Mar 26, 2016
@rcurtin rcurtin added this to the mlpack 2.0.2 milestone Mar 26, 2016
rcurtin commented Jun 8, 2016

Fixed in 428191f, 1c87ed9, c5b7186, and bc3916a.

@rcurtin rcurtin closed this Jun 8, 2016
@rcurtin rcurtin added the R: fixed label Jun 8, 2016
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment