CombinedObjectiveFunction.h

Go to the documentation of this file.

//===========================================================================
/*!
 * 
 *
 * \brief       CombinedObjectiveFunction
 * 
 * 
 *
 * \author      T.Voss, T. Glasmachers, O.Krause
 * \date        2010-2011
 *
 *
 * \par Copyright 1995-2017 Shark Development Team
 * 
 * <BR><HR>
 * This file is part of Shark.
 * <https://shark-ml.github.io/Shark/>
 * 
 * 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_COMBINEDOBJECTIVEFUNCTION_H
#define SHARK_OBJECTIVEFUNCTIONS_COMBINEDOBJECTIVEFUNCTION_H
 
 
#include <shark/ObjectiveFunctions/AbstractObjectiveFunction.h>
 
namespace shark {
 
 
///
/// \brief Linear combination of objective functions
///
/// \par
/// The CombinedObjectiveFunction is a linear combination of
/// objective functions. It assumed that the result type is
/// capable of forming linear combinations with real coefficients.
/// \ingroup objfunctions
template <typename SearchPointType, typename ResultT>
class CombinedObjectiveFunction : public AbstractObjectiveFunction<SearchPointType, ResultT>
{
public:
 
    typedef AbstractObjectiveFunction<SearchPointType, ResultT> super;
    typedef AbstractObjectiveFunction<SearchPointType, ResultT> element;
 
    /// Constructor
    CombinedObjectiveFunction(){
        this->m_features|=super::HAS_FIRST_DERIVATIVE;
        this->m_features|=super::HAS_SECOND_DERIVATIVE;
    }
 
    /// \brief From INameable: return the class name.
    std::string name() const
    { return "CombinedObjectiveFunction"; }
 
 
    /// Adds a new objective function with a
    /// weight of one to the linear combination.
    void add(element& e){
        add(1.0, e);
    }
 
    /// Adds a new objective function with
    /// a weight to the linear combination.
    void add(double weight, element& e)
    {
        SHARK_RUNTIME_CHECK(weight >= 0.0, "[CombinedObjectiveFunction::add] weight must be non-negative");
 
        m_weight.push_back(weight);
        m_elements.push_back(&e);
 
        if (e.features().test(element::IS_CONSTRAINED_FEATURE)) this->m_features.set(super::IS_CONSTRAINED_FEATURE);
        if (! e.features().test(element::HAS_FIRST_DERIVATIVE)) this->m_features.reset(super::HAS_FIRST_DERIVATIVE);
        if (! e.features().test(element::HAS_SECOND_DERIVATIVE)) this->m_features.reset(super::HAS_SECOND_DERIVATIVE);
    }
 
    /// Tests whether a point in SearchSpace is feasible,
    /// e.g., whether the constraints are fulfilled.
    bool isFeasible( const typename super::SearchPointType & input) const {
        std::size_t ic = m_elements.size();
        for ( std::size_t i=0; i<ic; i++)
            if (! m_elements[i]->isFeasible(input))
                return false;
        return true;
    }
    
    void init(){
        for ( std::size_t i=0; i<m_elements.size(); i++){
            m_elements[i]->setRng(this->mep_rng);
            m_elements[i]->init();
        }
    }
    
    std::size_t numberOfVariables()const{
        //todo sthis will fail if SarchPointType != Vectorspace
        return m_elements.size() == 0? 0: m_elements[0]->numberOfVariables();
    }
 
    /// Evaluates the objective function.
    typename super::ResultType eval( const typename super::SearchPointType & input ) const
    {
        ++this->m_evaluationCounter;
        std::size_t ic = m_elements.size();
        typename super::ResultType ret = m_weight[0] * m_elements[0]->eval(input);
        for (std::size_t i=1; i<ic; i++)
            ret += m_weight[i] * m_elements[i]->eval(input);
        return ret;
    }
 
    /// Evaluates the objective function
    /// and calculates its gradient.
    typename super::ResultType evalDerivative( const typename super::SearchPointType & input, typename super::FirstOrderDerivative & derivative ) const {
        ++this->m_evaluationCounter;
        SHARK_RUNTIME_CHECK(this->m_features.test(super::HAS_FIRST_DERIVATIVE), "[CombinedObjectiveFunction::evalDerivative] At least one of the objective functions combined is not differentiable");
        typename super::FirstOrderDerivative der;
        std::size_t ic = m_elements.size();
        typename super::ResultType ret = m_weight[0] * m_elements[0]->evalDerivative(input, der);
        derivative = m_weight[0] * der;
        for (std::size_t i=1; i != ic; i++)
        {
            ret += m_weight[i] * m_elements[i]->evalDerivative(input, der);
            derivative += m_weight[i] * der;
        }
        return ret;
    }
 
    /// Evaluates the objective function
    /// and calculates its gradient and
    /// its Hessian.
    typename super::ResultType evalDerivative( const typename super::SearchPointType & input, typename super::SecondOrderDerivative & derivative )const {
        SHARK_RUNTIME_CHECK(this->m_features.test(super::HAS_SECOND_DERIVATIVE), "[CombinedObjectiveFunction::evalDerivative] At least one of the objective functions combined is not twice differentiable");
        typename super::SecondOrderDerivative der;
        std::size_t ic = m_elements.size();
        typename super::ResultType ret = m_weight[0] * m_elements[0]->evalDerivative(input, der);
        derivative.gradient = m_weight[0] * der.gradient;
        derivative.hessian = m_weight[0] * der.hessian;
        for (std::size_t i=1; i<ic; i++)
        {
            ret += m_weight[i] * m_elements[i]->evalDerivative(input, der);
            derivative.gradient += m_weight[i] * der.gradient;
            derivative.hessian += m_weight[i] * der.hessian;
        }
        return ret;
    }
 
protected:
    /// list of weights
    std::vector<double> m_weight;
 
    /// list of "base" objective functions
    std::vector<element*> m_elements;
};
 
 
}
#endif // SHARK_CORE_COBINEDOBJECTIVEFUNCTION_H