Information gain normalized by the sum of the classification and split entropies see formula (11) in [Wehenkel1991] and appendix A in [Geurts2006]. More...

#include <score.h>

Inheritance diagram for curfil::NormalizedInformationGainScore:

Static Public Member Functions
template<class W >
__host__ static __device__ ScoreType	calculateScore (const size_t numClasses, const W leftClasses, const W rightClasses, const unsigned int leftRightStride, const W *allClasses, const ScoreType totalLeft, const ScoreType totalRight)

Static Protected Member Functions
__host__ static __device__ ScoreType	splitEntropy (const ScoreType total, const ScoreType totalLeft, const ScoreType totalRight)
template<class W >
__host__ static __device__ ScoreType	classificationEntropy (const size_t numLabels, const W *allClasses, const ScoreType total)
Static Protected Member Functions inherited from curfil::InformationGainScore
__host__ static __device__ ScoreType	entropy (const ScoreType prob)
__host__ static __device__ ScoreType	normalizeScore (const ScoreType score)

Detailed Description

Information gain normalized by the sum of the classification and split entropies see formula (11) in [Wehenkel1991] and appendix A in [Geurts2006].

Definition at line 120 of file score.h.

Member Function Documentation

template<class W >

__host__ static __device__ ScoreType curfil::NormalizedInformationGainScore::calculateScore	(	const size_t	numLabels,
		const W *	leftClasses,
		const W *	rightClasses,
		const unsigned int	leftRightStride,
		const W *	allClasses,
		const ScoreType	totalLeft,
		const ScoreType	totalRight
	)

inlinestatic

Returns: normalized score calculated after splitting

Reimplemented from curfil::InformationGainScore.

Definition at line 166 of file score.h.

template<class W >

__host__ static __device__ ScoreType curfil::NormalizedInformationGainScore::classificationEntropy	(	const size_t	numLabels,
		const W *	allClasses,
		const ScoreType	total
	)

inlinestaticprotected

Returns: classification entropy

Definition at line 141 of file score.h.

__host__ static __device__ ScoreType curfil::NormalizedInformationGainScore::splitEntropy	(	const ScoreType	total,
		const ScoreType	totalLeft,
		const ScoreType	totalRight
	)

inlinestaticprotected

Returns: H_s: "split entropy", i.e. a measure of the split balancedness

Definition at line 127 of file score.h.

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

src/curfil/score.h

Static Public Member Functions

Static Protected Member Functions

Detailed Description

Member Function Documentation