BIE::KSDistanceHeuristic Class Reference

KS Distance based heuristic. More...

#include <FrontierExpansionHeuristic.h>

Inheritance diagram for BIE::KSDistanceHeuristic:

[legend]

Collaboration diagram for BIE::KSDistanceHeuristic:

[legend]

List of all members.

Public Member Functions
	KSDistanceHeuristic (BaseDataTree *fulldist, double threshold)
	Constructor.
virtual bool	expandTile (vector< int > children, int parent)
	Expand the frontier by moving it from the current tile to its children.
Friends
class	boost::serialization::access

---

Detailed Description

KS Distance based heuristic.

This heuristic computes the KS statistic between each pair of child distributions proposed for expansion. The KS statistic gives the probability that two distributions come from the same underlying distribution If the KS statistic for any pair of child distributions is lower than the threshold, then the method recommends increasing the resolution of the tile, because their is sufficient evidence supporting the hypothesis that the distributions are different (and if they are different we want to look more closely). Setting a threshold at the 1% level would probably be pessimistic, since in this case expansion would not be recommended unless there was very good statistical evidence suggesting the distributions are different (from the rejection of the null hypothesis). The question of what a good value would be is still open.

---

The documentation for this class was generated from the following file:

• /home/weinberg/src/BIE/include/FrontierExpansionHeuristic.h