Line data    Source code

       1             : #pragma once
       2             : 
       3             : #include "Solver.h"
       4             : #include "Problem.h"
       5             : 
       6             : namespace elsa
       7             : {
       8             :     /**
       9             :      * @brief Class representing a simple gradient descent solver with a fixed, given step size.
      10             :      *
      11             :      * This class implements a simple gradient descent iterative solver with a fixed, given step
      12             :      * size. No particular stopping rule is currently implemented (only a fixed number of
      13             :      * iterations, default to 100).
      14             :      *
      15             :      * @tparam data_t data type for the domain and range of the problem, defaulting to real_t
      16             :      *
      17             :      * @see \verbatim embed:rst
      18             :      For a basic introduction and problem statement of first-order methods, see
      19             :      :ref:`here <elsa-first-order-methods-doc>` \endverbatim
      20             :      *
      21             :      * @author
      22             :      * - Tobias Lasser - initial code
      23             :      */
      24             :     template <typename data_t = real_t>
      25             :     class GradientDescent : public Solver<data_t>
      26             :     {
      27             :     public:
      28             :         /// Scalar alias
      29             :         using Scalar = typename Solver<data_t>::Scalar;
      30             : 
      31             :         /**
      32             :          * @brief Constructor for gradient descent, accepting a problem and a fixed step size
      33             :          *
      34             :          * @param[in] problem the problem that is supposed to be solved
      35             :          * @param[in] stepSize the fixed step size to be used while solving
      36             :          */
      37             :         GradientDescent(const Problem<data_t>& problem, data_t stepSize);
      38             : 
      39             :         /**
      40             :          * @brief Constructor for gradient descent, accepting a problem. The step size will be
      41             :          * computed as \f$ 1 \over L \f$ with \f$ L \f$ being the Lipschitz constant of the
      42             :          * function.
      43             :          *
      44             :          * @param[in] problem the problem that is supposed to be solved
      45             :          */
      46             :         GradientDescent(const Problem<data_t>& problem);
      47             : 
      48             :         /// make copy constructor deletion explicit
      49             :         GradientDescent(const GradientDescent<data_t>&) = delete;
      50             : 
      51             :         /// default destructor
      52          14 :         ~GradientDescent() override = default;
      53             : 
      54             :     private:
      55             :         /// the differentiable optimizaion problem
      56             :         std::unique_ptr<Problem<data_t>> _problem;
      57             : 
      58             :         /// the step size
      59             :         data_t _stepSize;
      60             : 
      61             :         /// the default number of iterations
      62             :         const index_t _defaultIterations{100};
      63             : 
      64             :         /**
      65             :          * @brief Solve the optimization problem, i.e. apply iterations number of iterations of
      66             :          * gradient descent
      67             :          *
      68             :          * @param[in] iterations number of iterations to execute (the default 0 value executes
      69             :          * _defaultIterations of iterations)
      70             :          *
      71             :          * @returns a reference to the current solution
      72             :          */
      73             :         DataContainer<data_t>& solveImpl(index_t iterations) override;
      74             : 
      75             :         /// implement the polymorphic clone operation
      76             :         GradientDescent<data_t>* cloneImpl() const override;
      77             : 
      78             :         /// implement the polymorphic comparison operation
      79             :         bool isEqual(const Solver<data_t>& other) const override;
      80             :     };
      81             : } // namespace elsa