AbstractLineSearchOptimizer.h

Go to the documentation of this file.

//===========================================================================
/*!
 * 
 *
 * \brief       Base class for Line Search Optimizer
 * \file
 * 
 *
 * \author      O. Krause
 * \date        2013
 *
 *
 * \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_ALGORITHMS_GRADIENTDESCENT_ABSTRACTLINESEARCHOPTIMIZER_H
#define SHARK_ALGORITHMS_GRADIENTDESCENT_ABSTRACTLINESEARCHOPTIMIZER_H
 
#include <shark/Algorithms/AbstractSingleObjectiveOptimizer.h>
#include <shark/Algorithms/GradientDescent/LineSearch.h>
 
namespace shark {
 
/// \brief Basis class for line search methods.
///
/// Line Search optimizer find an iterative optimum by starting from some point, choosing a search direction and than
/// performing a line search in that direction. To choose the search direction a local model of the function is  often used.
/// This class is a base class for all line search method which implement the general behaviour of line search methods.
/// Derived classes only need to implement initModel() and computeSearchDirection() to initializee and update
/// the model and find a new line search direction. The remaining functionality is implemented by the optimizer.
///
/// Also derived classes should specialise read() and write() methods for serialization if they have additional members
/// as well as choose a name() for the optimizer.
/// \ingroup gradientopt
template<class SearchPointType>
class AbstractLineSearchOptimizer : public AbstractSingleObjectiveOptimizer< SearchPointType > {
public:
    typedef typename AbstractSingleObjectiveOptimizer< SearchPointType >::ObjectiveFunctionType ObjectiveFunctionType;
protected:
    using AbstractSingleObjectiveOptimizer< SearchPointType >::m_best;
    /// \brief Initializes the internal model.
    ///
    /// Line Search Methods use a Model to search for the next search direction.
    /// The model is initialized during init()
    virtual void initModel() = 0;
 
    /// \brief Updates the Model and computes the next search direction
    ///
    /// After a step was performed, this method is called to compute the next 
    /// search direction. This usually involves updating the internal model using the 
    /// new and old step information. Afterwards m_searchDirection should contain
    /// the next search direction.
    virtual void computeSearchDirection(ObjectiveFunctionType const& objectiveFunction) = 0;
 
public:
    AbstractLineSearchOptimizer();
 
    void init(ObjectiveFunctionType const& objectiveFunction,  SearchPointType const& startingPoint) ;
    
    using AbstractSingleObjectiveOptimizer< SearchPointType >::init;
 
    void step(ObjectiveFunctionType const& objectiveFunction);
 
    //from ISerializable
    void read(InArchive &archive);
    void write(OutArchive &archive) const;
 
 
    //linesearch handling
    LineSearch<SearchPointType> const& lineSearch()const {
        return m_linesearch;
    }
    LineSearch<SearchPointType>& lineSearch() {
        return m_linesearch;
    }
    
    /// \brief Returns the derivative at the current point. Can be used for stopping criteria.
    SearchPointType const& derivative()const{
        return m_derivative;
    }
 
 
protected: // Instance vars
 
    LineSearch<SearchPointType> m_linesearch; ///< used line search method.
    std::size_t m_dimension; ///< number of parameters
    double m_initialStepLength;///< Initial step length to begin with the line search.
 
    SearchPointType  m_derivative; ///< gradient of m_best.point
    SearchPointType  m_searchDirection;///< search direction of next step
 
    //information from previous step
    SearchPointType m_lastPoint; ///<  previous point
    SearchPointType m_lastDerivative; ///< gradient of the previous point
    double m_lastValue;     ///< value of the previous point
};
 
extern template class AbstractLineSearchOptimizer<RealVector>;
extern template class AbstractLineSearchOptimizer<FloatVector>;
#ifdef SHARK_USE_OPENCL
extern template class AbstractLineSearchOptimizer<RealGPUVector>;
extern template class AbstractLineSearchOptimizer<FloatGPUVector>;
#endif
}
#endif