stochastic_gradient_descent.h

//   OpenNN: Open Neural Networks Library
//   www.opennn.net
//
//   S T O C H A S T I C   G R A D I E N T   D E S C E N T   C L A S S   H E A D E R
//
//   Artificial Intelligence Techniques SL
//   artelnics@artelnics.com
 
#ifndef STOCHASTICGRADIENTDESCENT_H
#define STOCHASTICGRADIENTDESCENT_H
 
// System includes
 
#include <string>
#include <sstream>
#include <iostream>
#include <fstream>
#include <algorithm>
#include <functional>
#include <limits>
#include <cmath>
#include <ctime>
 
// OpenNN includes
 
#include "config.h"
 
#include "loss_index.h"
#include "optimization_algorithm.h"
 
namespace OpenNN
{
 
struct StochasticGradientDescentData;
 
 
 
class StochasticGradientDescent : public OptimizationAlgorithm
{
 
public:
 
   // Constructors
 
   explicit StochasticGradientDescent(); 
 
   explicit StochasticGradientDescent(LossIndex*);
 
   // Destructor
 
   virtual ~StochasticGradientDescent();
   
   //Training operators
 
   const type& get_initial_learning_rate() const;
   const type& get_initial_decay() const;
   const type& get_momentum() const;
   const bool& get_nesterov() const;
 
   // Stopping criteria
 
   const type& get_loss_goal() const;
   const type& get_maximum_time() const;
 
   // Set methods
 
   void set_loss_index_pointer(LossIndex*);
 
   void set_default();
 
   void set_batch_samples_number(const Index& new_batch_samples_number)
   {
       batch_samples_number = new_batch_samples_number;
   }
 
   // Get methods
 
   Index get_batch_samples_number() const;
 
   //Training operators
 
   void set_initial_learning_rate(const type&);
   void set_initial_decay(const type&);
   void set_momentum(const type&);
   void set_nesterov(const bool&);
 
   void set_maximum_epochs_number(const Index&);
 
   // Stopping criteria
 
   void set_loss_goal(const type&);
   void set_maximum_time(const type&);
 
   // Training methods
 
   void update_parameters(LossIndexBackPropagation& back_propagation,
                         StochasticGradientDescentData& optimization_data);
 
   TrainingResults perform_training();
 
   string write_optimization_algorithm_type() const;
 
   // Serialization methods
 
   Tensor<string, 2> to_string_matrix() const;
 
   void from_XML(const tinyxml2::XMLDocument&);
 
   void write_XML(tinyxml2::XMLPrinter&) const;   
 
private:
 
   // Training operators
 
 
   type initial_learning_rate;
 
 
   type initial_decay;
 
 
   type momentum;
 
 
   bool nesterov;
 
 
   Index batch_samples_number = 1000;
 
   // Stopping criteria
 
 
   type training_loss_goal = type(0);
 
 
   Index maximum_selection_failures = numeric_limits<Index>::max();
 
 
   Index maximum_epochs_number = 10000;
 
 
   type maximum_time = type(3600);
 
#ifdef OPENNN_CUDA
    #include "../../opennn-cuda/opennn-cuda/stochastic_gradient_descent_cuda.h"
#endif
 
};
 
 
struct StochasticGradientDescentData : public OptimizationAlgorithmData
{
 
    explicit StochasticGradientDescentData()
    {
    }
 
    explicit StochasticGradientDescentData(StochasticGradientDescent* new_stochastic_gradient_descent_pointer)
    {
        set(new_stochastic_gradient_descent_pointer);
    }
 
    virtual ~StochasticGradientDescentData() {}
 
    void set(StochasticGradientDescent* new_stochastic_gradient_descent_pointer)
    {
        stochastic_gradient_descent_pointer = new_stochastic_gradient_descent_pointer;
 
        LossIndex* loss_index_pointer = stochastic_gradient_descent_pointer->get_loss_index_pointer();
 
        NeuralNetwork* neural_network_pointer = loss_index_pointer->get_neural_network_pointer();
 
        const Index parameters_number = neural_network_pointer->get_parameters_number();
 
        parameters_increment.resize(parameters_number);
        nesterov_increment.resize(parameters_number);
        last_parameters_increment.resize(parameters_number);
 
        parameters_increment.setZero();
        nesterov_increment.setZero();
        last_parameters_increment.setZero();
    }
 
    StochasticGradientDescent* stochastic_gradient_descent_pointer = nullptr;
 
    Index iteration = 0;
 
    Tensor<type, 1> parameters_increment;
    Tensor<type, 1> nesterov_increment;
    Tensor<type, 1> last_parameters_increment;
};
 
}
 
#endif