AbstractOptimizer.h

Go to the documentation of this file.

//===========================================================================
/*!
 * 
 *
 * \brief       AbstractOptimizer
 * 
 * 
 *
 * \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.
 * <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_ABSTRACTOPTIMIZER_H
#define SHARK_OBJECTIVEFUNCTIONS_ABSTRACTOPTIMIZER_H

#include <shark/ObjectiveFunctions/AbstractObjectiveFunction.h>

namespace shark {

/// \brief An optimizer that optimizes general objective functions
/// 
/// After construction and configurationg the optimizer, init() is called with the objective function
/// to be used. After that step() can be called until the required solution is found. The solution can be queried
/// using solution(). The type of the solution depends on the optimisation problem at hand.
/// It is allowed to add constrains on the features the objective function needs to offer
/// 
/// These are:
/// - REQUIRES_VALUE: The function is evaluated to use the optimizer and
///   the HAS_VALUE-flag must be set
/// - REQUIRES_FIRST_DERIVATIVE: The first derivative needs to be evaluated and
/// - HAS_FIRST_DERIVATIVE must be set
/// - REQUIRES_SECOND_DERIVATIVE: The second derivative needs to be evaluated and
/// - HAS_SECOND_DERIVATIVE must be set
/// - CAN_SOLVE_CONSTRAINED: The optimizer can solve functions which are constrained and
///   where the IS_CONSTRAINED_FEATURE is set.
/// - REQUIRES_CLOSEST_FEASIBLE: If the function is constrained, it must offer a way to
/// construct the closest feasible point and
/// - CAN_PROVIDE_CLOSEST_FEASIBLE must be set
/// 
/// Also when init() is called as offered by the AbstractOptimizer interface, the function
/// is required to have the CAN_PROPOSE_STARTING_POINT flag.
/// 
/// \tparam PointType The type of search space the optimizer works upon.
/// \tparam ResultT The objective space the optimizer works upon.
/// \tparam SolutionTypeT The type of the final solution.
template <typename PointType, typename ResultT, typename SolutionTypeT>
class AbstractOptimizer : public INameable, public ISerializable {
public:
    typedef PointType SearchPointType;
    typedef ResultT ResultType;
    typedef SolutionTypeT SolutionType;
    typedef AbstractObjectiveFunction<PointType,ResultType> ObjectiveFunctionType;

    /// \brief Models features that the optimizer requires from the objective function.
    /// \sa AbstractObjectiveFunction
    enum Feature {
        REQUIRES_VALUE              =  1,
        REQUIRES_FIRST_DERIVATIVE   =  2,
        REQUIRES_SECOND_DERIVATIVE  =  4,
        CAN_SOLVE_CONSTRAINED           =  8,
        REQUIRES_CLOSEST_FEASIBLE       = 16
    };

    SHARK_FEATURE_INTERFACE;
    
    bool requiresValue()const{
        return features()& REQUIRES_VALUE;
    }
    
    bool requiresFirstDerivative()const{
        return features()& REQUIRES_FIRST_DERIVATIVE;
    }
    bool requiresSecondDerivative()const{
        return features()& REQUIRES_SECOND_DERIVATIVE;
    }
    bool canSolveConstrained()const{
        return features()& CAN_SOLVE_CONSTRAINED;
    }
    bool requiresClosestFeasible()const{
        return features()& REQUIRES_CLOSEST_FEASIBLE;
    }

    virtual ~AbstractOptimizer() {}
    
    /// \brief Returns the number of points this method requires for initialisation
    ///
    /// The number of points supplied is allowed to be smaller, however in this case
    /// the optimizer will resort to techniques generating additional points if needed.
    virtual std::size_t numInitPoints() const = 0;

    /// \brief Initialize the optimizer for the supplied objective function.
    ///
    /// Be aware that function.init() has to be called before calling this function!
    /// This function will initialize the algorithm with a number of points proposed
    /// by the function. Note that this must fail if the function can not propose starting point(s).
    ///
    /// \param [in] function The objective function to initialize for.
    virtual void init( ObjectiveFunctionType const& function ){
        SHARK_RUNTIME_CHECK(function.canProposeStartingPoint(), "Objective function does not propose a starting point");
        std::vector<SearchPointType> initPoints(numInitPoints());
        for(SearchPointType& point: initPoints){
            point = function.proposeStartingPoint();
        }
        init(function,initPoints);
    }
    
    
    /// \brief Initialize the optimizer for the supplied objective function using a set of initialisation points
    ///
    /// Most single objective algorithms only require a single point. However multi-objective algorithms
    /// need a set of initialisation points. The number of points required should be numInitPoints().
    /// Otherwise the algorithm might use heuristics to generate additional points if needed.
    ///
    /// Be aware that function.init() has to be called before calling this function!
    /// \param [in] function The objective function to initialize for.
    /// \param [in] initPoints points used for initialisation. Should be at least numInitPoints().
    virtual void init( ObjectiveFunctionType const& function, std::vector<SearchPointType> const& initPoints ) = 0;

    /// \brief Carry out one step of the optimizer for the supplied objective function.
    /// \param [in] function The objective function to initialize for.
    virtual void step( ObjectiveFunctionType const& function ) = 0;

    /// \brief Accesses the best solution obtained so far. 
    /// \return An immutable reference to the best solution obtained so far.
    virtual SolutionType const& solution() const = 0; //mt_hint: try accessing this thing via solution().point and solution().value..

protected:
    /// \brief Convenience function that checks whether the features of the supplied objective function match with the required features of the optimizer.
    /// \param [in] objectiveFunction The function to match with.
    /// \throws shark::Exception
    void checkFeatures (ObjectiveFunctionType const& objectiveFunction){
        //test whether the function can be evaluated
        SHARK_RUNTIME_CHECK(!requiresValue() || objectiveFunction.hasValue(), name()+" Requires value of objective function");
        //test first derivative
        SHARK_RUNTIME_CHECK(!requiresFirstDerivative() || objectiveFunction.hasFirstDerivative(), name()+" Requires first derivative of objective function");
        //test second derivative
        SHARK_RUNTIME_CHECK(!requiresSecondDerivative() || objectiveFunction.hasSecondDerivative(), name()+" Requires second derivative of objective function");
        //test for constraints
        if(objectiveFunction.isConstrained()){
            SHARK_RUNTIME_CHECK(canSolveConstrained(), name()+" Can not solve constrained problems");
            SHARK_RUNTIME_CHECK(
                !requiresClosestFeasible() || objectiveFunction.canProvideClosestFeasible(), 
                name()+" Requires closest feasible solution for solving constrained problems"
            );
        }
    }
};

}

#endif // SHARK_CORE_ABSTRACTOPTIMIZER_H