parametric_relu.hpp

Go to the documentation of this file.

#ifndef MLPACK_METHODS_ANN_LAYER_PReLU_HPP
#define MLPACK_METHODS_ANN_LAYER_PReLU_HPP

#include <mlpack/prereqs.hpp>

namespace mlpack {
namespace ann  {

template <
    typename InputDataType = arma::mat,
    typename OutputDataType = arma::mat
>
class PReLU
{
 public:
  PReLU(const double userAlpha = 0.03);

  /*
   * Reset the layer parameter.
   */
  void Reset();

  template<typename InputType, typename OutputType>
  void Forward(const InputType&& input, OutputType&& output);

  template<typename DataType>
  void Backward(const DataType&& input, DataType&& gy, DataType&& g);

  template<typename eT>
  void Gradient(const arma::Mat<eT>&& input,
                arma::Mat<eT>&& error,
                arma::Mat<eT>&& gradient);

  OutputDataType const& Parameters() const { return alpha; }
  OutputDataType& Parameters() { return alpha; }

  InputDataType const& InputParameter() const { return inputParameter; }
  InputDataType& InputParameter() { return inputParameter; }

  OutputDataType const& OutputParameter() const { return outputParameter; }
  OutputDataType& OutputParameter() { return outputParameter; }

  OutputDataType const& Delta() const { return delta; }
  OutputDataType& Delta() { return delta; }

  OutputDataType const& Gradient() const { return gradient; }
  OutputDataType& Gradient() { return gradient; }

  double const& Alpha() const { return alpha(0); }
  double& Alpha() { return alpha(0); }

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

 private:
  double Fn(const double x)
  {
    return std::max(x, alpha(0) * x);
  }

  template<typename eT>
  void Fn(const arma::Mat<eT>& x, arma::Mat<eT>& y)
  {
    y = x;
    arma::uvec negative = arma::find(x < 0);
    y(negative) = x(negative) * alpha(0);
  }

  double Deriv(const double x)
  {
    return (x >= 0) ? 1 : alpha(0);
  }

  template<typename InputType, typename OutputType>
  void Deriv(const InputType& x, OutputType& y)
  {
    y = x;

    for (size_t i = 0; i < x.n_elem; i++)
    {
      y(i) = Deriv(x(i));
    }
  }

  OutputDataType delta;

  InputDataType inputParameter;

  OutputDataType outputParameter;

  OutputDataType alpha;

  OutputDataType gradient;

  double userAlpha;
}; // class PReLU

} // namespace ann
} // namespace mlpack

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

#endif