rnn.hpp

Go to the documentation of this file.

#ifndef MLPACK_METHODS_ANN_RNN_HPP
#define MLPACK_METHODS_ANN_RNN_HPP

#include <mlpack/prereqs.hpp>

#include "visitor/delete_visitor.hpp"
#include "visitor/delta_visitor.hpp"
#include "visitor/output_parameter_visitor.hpp"
#include "visitor/reset_visitor.hpp"
#include "visitor/weight_size_visitor.hpp"

#include <mlpack/methods/ann/layer/layer_types.hpp>
#include <mlpack/methods/ann/init_rules/random_init.hpp>
#include <mlpack/core/optimizers/sgd/sgd.hpp>

namespace mlpack {
namespace ann  {

template<
  typename OutputLayerType = NegativeLogLikelihood<>,
  typename InitializationRuleType = RandomInitialization
>
class RNN
{
 public:
  using NetworkType = RNN<OutputLayerType, InitializationRuleType>;

  RNN(const size_t rho,
      const bool single = false,
      OutputLayerType outputLayer = OutputLayerType(),
      InitializationRuleType initializeRule = InitializationRuleType());

  RNN(const arma::mat& predictors,
      const arma::mat& responses,
      const size_t rho,
      const bool single = false,
      OutputLayerType outputLayer = OutputLayerType(),
      InitializationRuleType initializeRule = InitializationRuleType());

  ~RNN();

  template<
      template<typename> class OptimizerType = mlpack::optimization::SGD
  >
  void Train(const arma::mat& predictors,
             const arma::mat& responses,
             OptimizerType<NetworkType>& optimizer);

  template<
      template<typename> class OptimizerType = mlpack::optimization::SGD
  >
  void Train(const arma::mat& predictors, const arma::mat& responses);

  void Predict(arma::mat& predictors, arma::mat& responses);

  double Evaluate(const arma::mat& /* parameters */,
                  const size_t i,
                  const bool deterministic = true);

  void Gradient(const arma::mat& parameters,
                const size_t i,
                arma::mat& gradient);

  /*
   * Add a new module to the model.
   *
   * @param layer The Layer to be added to the model.
   */
  template<typename LayerType>
  void Add(const LayerType& layer) { network.push_back(new LayerType(layer)); }

  /*
   * Add a new module to the model.
   *
   * @param args The layer parameter.
   */
  template <class LayerType, class... Args>
  void Add(Args... args) { network.push_back(new LayerType(args...)); }

  /*
   * Add a new module to the model.
   *
   * @param layer The Layer to be added to the model.
   */
  void Add(LayerTypes layer) { network.push_back(layer); }

  size_t NumFunctions() const { return numFunctions; }

  const arma::mat& Parameters() const { return parameter; }
  arma::mat& Parameters() { return parameter; }

  template<typename Archive>
  void Serialize(Archive& ar, const unsigned int /* version */);

 private:
  // Helper functions.
  void Forward(arma::mat&& input);

  void Backward();

  void Gradient();

  /*
   * Predict the response of the given input sequence.
   *
   * @param predictors Input predictors.
   * @param responses Vector to put output prediction of a response into.
   */
  void SinglePredict(const arma::mat& predictors, arma::mat& responses);

  void ResetParameters();

  void ResetDeterministic();

  void ResetGradients(arma::mat& gradient);

  size_t rho;

  OutputLayerType outputLayer;

  InitializationRuleType initializeRule;

  size_t inputSize;

  size_t outputSize;

  size_t targetSize;

  bool reset;

  bool single;

  std::vector<LayerTypes> network;

  arma::mat predictors;

  arma::mat responses;

  arma::mat parameter;

  size_t numFunctions;

  arma::mat error;

  arma::mat currentInput;

  DeltaVisitor deltaVisitor;

  OutputParameterVisitor outputParameterVisitor;

  std::vector<arma::mat> moduleOutputParameter;

  WeightSizeVisitor weightSizeVisitor;

  ResetVisitor resetVisitor;

  DeleteVisitor deleteVisitor;

  bool deterministic;
}; // class RNN

} // namespace ann
} // namespace mlpack

// Include implementation.
#include "rnn_impl.hpp"

#endif