parametric_relu_impl.hpp

Go to the documentation of this file.

#ifndef MLPACK_METHODS_ANN_LAYER_PReLU_IMPL_HPP
#define MLPACK_METHODS_ANN_LAYER_PReLU_IMPL_HPP

// In case it hasn't yet been included.
#include "parametric_relu.hpp"

namespace mlpack {
namespace ann  {

template<typename InputDataType, typename OutputDataType>
PReLU<InputDataType, OutputDataType>::PReLU(
    const double userAlpha) : userAlpha(userAlpha)
{
  alpha.set_size(WeightSize(), 1);
  alpha(0) = userAlpha;
}

template<typename InputDataType, typename OutputDataType>
void PReLU<InputDataType, OutputDataType>::Reset()
{
  alpha(0) = userAlpha;
}

template<typename InputDataType, typename OutputDataType>
template<typename InputType, typename OutputType>
void PReLU<InputDataType, OutputDataType>::Forward(
    const InputType& input, OutputType& output)
{
  output = input;
  arma::uvec negative = arma::find(input < 0);
  output(negative) = input(negative) * alpha(0);
}

template<typename InputDataType, typename OutputDataType>
template<typename DataType>
void PReLU<InputDataType, OutputDataType>::Backward(
    const DataType& input, const DataType& gy, DataType& g)
{
  DataType derivative;
  derivative.set_size(arma::size(input));
  for (size_t i = 0; i < input.n_elem; ++i)
  {
    derivative(i) = (input(i) >= 0) ? 1 : alpha(0);
  }

  g = gy % derivative;
}

template<typename InputDataType, typename OutputDataType>
template<typename eT>
void PReLU<InputDataType, OutputDataType>::Gradient(
    const arma::Mat<eT>& input,
    const arma::Mat<eT>& error,
    arma::Mat<eT>& gradient)
{
  if (gradient.n_elem == 0)
  {
    gradient = arma::zeros<arma::mat>(1, 1);
  }

  arma::mat zeros = arma::zeros<arma::mat>(input.n_rows, input.n_cols);
  gradient(0) = arma::accu(error % arma::min(zeros, input)) / input.n_cols;
}

template<typename InputDataType, typename OutputDataType>
template<typename Archive>
void PReLU<InputDataType, OutputDataType>::serialize(
    Archive& ar,
    const uint32_t /* version */)
{
  ar(CEREAL_NVP(alpha));
}

} // namespace ann
} // namespace mlpack

#endif