glearner.h

来自「一个由Mike Gashler完成的机器学习方面的includes neural」· C头文件 代码 · 共 81 行

/*
	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 __GLEARNER_H__
#define __GLEARNER_H__

class GArffRelation;
class GArffData;


class GSupervisedLearner
{
protected:
	GArffRelation* m_pRelation;

public:
	GSupervisedLearner(GArffRelation* pRelation);
	virtual ~GSupervisedLearner();

	// Returns the relation used to construct this learner
	GArffRelation* GetRelation() { return m_pRelation; }

	// Discard any training (but not any settings) so it can be trained again
	virtual void Reset() = 0;

	// Train with the provided data
	virtual void Train(GArffData* pData) = 0;

	// Evaluates the input values in the provided vector and
	// deduce the output values
	virtual void Eval(double* pVector) = 0;

	// Computes predictive accuracy (the ratio of samples that
	// are correctly classified to total samples). If there is
	// more than one output attribute, each output attribute
	// is evaluated independently. If there are continuous output
	// values, it uses 1-1/(1+(squared error)) as an estimate
	// so that a small squared error will be close to 1 (correct)
	// and a large squared error will be close to 0 (incorrect).
	double MeasurePredictiveAccuracy(GArffData* pData);

	// Computes the mean squared error. If there are multiple
	// output attributes, each one is considered independently.
	// If there are discreet output attributes, a correct
	// classification is considered to be a squared error of
	// 0 and an incorrect classification is a squared error of 1.
	double MeasureMeanSquaredError(GArffData* pData);

	// Perform n-fold cross validation on pData. If bRegression is true,
	// it will return the average mean squared error. If bRegression is
	// false, it will return the average predictive accuracy.
	double CrossValidate(GArffData* pData, int nFolds, bool bRegression);
};


// Always outputs the mean (for continuous values) and the most common
// class (for enumerated values)
class GBaselineLearner : public GSupervisedLearner
{
protected:
	double* m_pOutputs;

public:
	GBaselineLearner(GArffRelation* pRelation);
	virtual ~GBaselineLearner();

	virtual void Reset();
	virtual void Train(GArffData* pData);
	virtual void Eval(double* pVector);
};

#endif // __GLEARNER_H__