//===========================================================================
/*!
 * 
 *
 * \brief       Flexible error measure for classication tasks
 * 
 * 
 *
 * \author      T. Glasmachers
 * \date        2011
 *
 *
 * \par Copyright 1995-2017 Shark Development Team
 * 
 * <BR><HR>
 * This file is part of Shark.
 * <http://shark-ml.org/>
 * 
 * Shark 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 3 of the License, or
 * (at your option) any later version.
 * 
 * Shark is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU Lesser General Public License for more details.
 * 
 * You should have received a copy of the GNU Lesser General Public License
 * along with Shark.  If not, see <http://www.gnu.org/licenses/>.
 *
 */

#ifndef SHARK_OBJECTIVEFUNCTIONS_LOSS_DISCRETELOSS_H
#define SHARK_OBJECTIVEFUNCTIONS_LOSS_DISCRETELOSS_H

#include <shark/Core/DLLSupport.h>
#include <shark/ObjectiveFunctions/Loss/AbstractLoss.h>


namespace shark {


///
/// \brief flexible loss for classification
///
/// \par
/// The DiscreteLoss class allows for the definition of
/// a cost matrix applied to a finite number of classes.
/// The cost of correct classification must be zero, all
/// other costs must be non-negative.
///
/// \par
/// Note: As a special case, this loss can be used to provide
/// a balanced error signal for unbalanced data sets.
///
class DiscreteLoss : public AbstractLoss<unsigned int, unsigned int>
{
public:

	/// Constructor
	/// \param  cost     cost matrix in the format (target, prediction).
	SHARK_EXPORT_SYMBOL DiscreteLoss(RealMatrix const& cost);


	/// \brief From INameable: return the class name.
	std::string name() const
	{ return "DiscreteLoss"; }

	/// inherited from AbstractLoss, evaluation of the loss function
	SHARK_EXPORT_SYMBOL double eval(BatchLabelType const& target, BatchOutputType const& prediction) const;

	/// Define a new cost structure given by an explicit cost matrix.
	/// \param  cost   cost matrix in the format (target, prediction).
	SHARK_EXPORT_SYMBOL void defineCostMatrix(RealMatrix const& cost);

	/// Define a new cost structure so that the cost of misclassifying
	/// a pattern is anti-proportional to the frequency of its class.
	/// This amounts to balancing the class-wise cost in unbalanced
	/// data sets (i.e., where one class is far more frequent than
	/// another).
	///
	/// \param  labels   label set to which the balanced loss should be adapted
	SHARK_EXPORT_SYMBOL void defineBalancedCost(UnlabeledData<unsigned int> const& labels);

protected:
	/// cost matrix
	RealMatrix m_cost;
};


}
#endif