gnaiveinstance.h

一个由Mike Gashler完成的机器学习方面的includes neural net, naive bayesian classifier, decision tree, KNN, a genet

/*
	Copyright (C) 2006, Mike Gashler

	This library is free software; you can redistribute it and/or
	modify it under the terms of the GNU Lesser General Public
	License as published by the Free Software Foundation; either
	version 2.1 of the License, or (at your option) any later version.

	see http://www.gnu.org/copyleft/lesser.html
*/

#ifndef __GNAIVEINSTANCE_H__
#define __GNAIVEINSTANCE_H__

#include "GLearner.h"

class GNaiveInstanceAttr;


// This is a cross between a KNN instance learner and a Naive Bayes learner.
// It's not yet well tested, but in theory it should be able to regress with
// precision like KNN, but be tolerant to the curse of dimensionality like
// Naive Bayes.
class GNaiveInstance : public GSupervisedLearner
{
protected:
	GNaiveInstanceAttr** m_pAttrs;
	int m_nNeighbors;
	double m_dMinDist;
	double* m_pValueSums;
	double* m_pWeightSums;
	double* m_pSumBuffer;
	double* m_pSumOfSquaresBuffer;
	double* m_pInputBuffer;
	int m_nVectorModeInputs;
	int m_nVectorModeOutputs;

public:
	// nNeighbors is the number of neighbors (in every dimension)
	// that will contribute to the output value. dDiscernment tells
	// it how much to emphasize low-variance patterns over high-variance
	// patterns. (0=none, 1=proportional to variance ratio)
	GNaiveInstance(GArffRelation* pRelation, int nNeighbors);
	virtual ~GNaiveInstance();

	// Discard any training (but not any settings) so it can be trained again
	virtual void Reset();

	// Incrementally train with a single instance
	void AddInstance(double* pVector);

	// Train using all the samples in a collection
	virtual void Train(GArffData* pData);

	// Deduce the output values from the input values
	virtual void Eval(double* pVector);

protected:
	void EvalInput(int nInputDim, double dInput);
};

#endif // __GNAIVEINSTANCE_H__