Package elki.index.tree.spatial.kd.split

Interface SplitStrategy

All Known Implementing Classes:

BoundedMidpointSplit, LeastOneDimSSQSplit, LeastSSQSplit, MeanVarianceSplit, MedianSplit, MedianVarianceSplit, MidpointSplit

public interface SplitStrategy

Split strategy for full k-d-tree construction. This interface is engineered to work on a modifiable array of DBIDs (more precisely, on a range), and the split partitions are encoded in this array.

Since:

0.8.0

Author:

Erich Schubert

Nested Class Summary

Nested Classes

Modifier and Type	Interface	Description
static class	SplitStrategy.Info	Split information.
static class	SplitStrategy.Util	Utility functions.

Method Summary

Modifier and Type	Method	Description
SplitStrategy.Info	findSplit(Relation<? extends NumberVector> relation, int dims, ArrayModifiableDBIDs sorted, DBIDArrayMIter iter, int left, int right, VectorUtil.SortDBIDsBySingleDimension comp)	Build the k-d-tree using midpoint splitting.

Method Detail

findSplit

SplitStrategy.Info findSplit(Relation<? extends NumberVector> relation,
                             int dims,
                             ArrayModifiableDBIDs sorted,
                             DBIDArrayMIter iter,
                             int left,
                             int right,
                             VectorUtil.SortDBIDsBySingleDimension comp)

Build the k-d-tree using midpoint splitting.

Parameters:

relation - data relation

dims - data dimensionality

sorted - modifiable array to encode the partitions

iter - reusable iterator to the array

left - left subinterval

right - right subinterval

comp - reusable comparator (bound to the relation above)

Returns:

split info, null if no good split