[mlpack-git] master: Force re-Precalculate() when we run out of points in one cluster. (ec8a151)
gitdub at big.cc.gt.atl.ga.us
gitdub at big.cc.gt.atl.ga.us
Mon Nov 23 08:26:04 EST 2015
Repository : https://github.com/mlpack/mlpack
On branch : master
Link : https://github.com/mlpack/mlpack/compare/1bb8434eec26338121f6bcfbbf7c83daff35a570...ec8a15165b12c246df58cb5e66afe1c6d865ce07
>---------------------------------------------------------------
commit ec8a15165b12c246df58cb5e66afe1c6d865ce07
Author: Ryan Curtin <ryan at ratml.org>
Date: Mon Nov 23 08:25:39 2015 -0500
Force re-Precalculate() when we run out of points in one cluster.
See #481.
>---------------------------------------------------------------
ec8a15165b12c246df58cb5e66afe1c6d865ce07
src/mlpack/methods/kmeans/max_variance_new_cluster_impl.hpp | 11 ++++++++++-
1 file changed, 10 insertions(+), 1 deletion(-)
diff --git a/src/mlpack/methods/kmeans/max_variance_new_cluster_impl.hpp b/src/mlpack/methods/kmeans/max_variance_new_cluster_impl.hpp
index fe82e5f..6ea9d0c 100644
--- a/src/mlpack/methods/kmeans/max_variance_new_cluster_impl.hpp
+++ b/src/mlpack/methods/kmeans/max_variance_new_cluster_impl.hpp
@@ -64,12 +64,21 @@ size_t MaxVarianceNewCluster::EmptyCluster(const MatType& data,
// Modify the variances, as necessary.
variances[emptyCluster] = 0;
- // One has already been subtracted from clusterCounts[maxVarCluster].
+ // One has already been subtracted from clusterCounts[maxVarCluster]. If
+ // EmptyCluster() is called again, we can't pull another point from
+ // maxVarCluster (we'd be making an empty cluster), so force a call to
+ // Precalculate() if EmptyCluster() is called again by changing
+ // this->iteration.
if (clusterCounts[maxVarCluster] <= 1)
+ {
variances[maxVarCluster] = 0;
+ --this->iteration;
+ }
else
+ {
variances[maxVarCluster] = (1.0 / clusterCounts[maxVarCluster]) *
((clusterCounts[maxVarCluster] + 1) * variances[maxVarCluster] - maxDistance);
+ }
// Output some debugging information.
Log::Debug << "Point " << furthestPoint << " assigned to empty cluster " <<
More information about the mlpack-git
mailing list