glorot_init.hpp

Go to the documentation of this file.

#ifndef MLPACK_METHODS_ANN_INIT_RULES_GLOROT_INIT_HPP
#define MLPACK_METHODS_ANN_INIT_RULES_GLOROT_INIT_HPP

#include <mlpack/prereqs.hpp>
#include "random_init.hpp"
#include "gaussian_init.hpp"

using namespace mlpack::math;

namespace mlpack {
namespace ann  {

template<bool Uniform = true>
class GlorotInitializationType
{
 public:
  GlorotInitializationType()
  {
    // Nothing to do here.
  }

  template<typename eT>
  void Initialize(arma::Mat<eT>& W,
                  const size_t rows,
                  const size_t cols);

  template<typename eT>
  void Initialize(arma::Mat<eT>& W);

  template<typename eT>
  void Initialize(arma::Cube<eT>& W,
                  const size_t rows,
                  const size_t cols,
                  const size_t slices);

  template<typename eT>
  void Initialize(arma::Cube<eT>& W);
}; // class GlorotInitializationType

template <>
template<typename eT>
inline void GlorotInitializationType<false>::Initialize(arma::Mat<eT>& W,
                                                       const size_t rows,
                                                       const size_t cols)
{
  if (W.is_empty())
    W.set_size(rows, cols);

  double var = 2.0 / double(rows + cols);
  GaussianInitialization normalInit(0.0, var);
  normalInit.Initialize(W, rows, cols);
}

template <>
template<typename eT>
inline void GlorotInitializationType<false>::Initialize(arma::Mat<eT>& W)
{
  if (W.is_empty())
    Log::Fatal << "Cannot initialize and empty matrix." << std::endl;

  double var = 2.0 / double(W.n_rows + W.n_cols);
  GaussianInitialization normalInit(0.0, var);
  normalInit.Initialize(W);
}

template <>
template<typename eT>
inline void GlorotInitializationType<true>::Initialize(arma::Mat<eT>& W,
                                                       const size_t rows,
                                                       const size_t cols)
{
  if (W.is_empty())
    W.set_size(rows, cols);

  // Limit of distribution.
  double a = sqrt(6) / sqrt(rows + cols);
  RandomInitialization randomInit(-a, a);
  randomInit.Initialize(W, rows, cols);
}

template <>
template<typename eT>
inline void GlorotInitializationType<true>::Initialize(arma::Mat<eT>& W)
{
  if (W.is_empty())
    Log::Fatal << "Cannot initialize an empty matrix." << std::endl;

  // Limit of distribution.
  double a = sqrt(6) / sqrt(W.n_rows + W.n_cols);
  RandomInitialization randomInit(-a, a);
  randomInit.Initialize(W);
}

template <bool Uniform>
template<typename eT>
inline void GlorotInitializationType<Uniform>::Initialize(arma::Cube<eT>& W,
                                                          const size_t rows,
                                                          const size_t cols,
                                                          const size_t slices)
{
  if (W.is_empty())
    W.set_size(rows, cols, slices);

  for (size_t i = 0; i < slices; ++i)
    Initialize(W.slice(i), rows, cols);
}

template <bool Uniform>
template<typename eT>
inline void GlorotInitializationType<Uniform>::Initialize(arma::Cube<eT>& W)
{
  if (W.is_empty())
    Log::Fatal << "Cannot initialize an empty matrix." << std::endl;

  for (size_t i = 0; i < W.n_slices; ++i)
    Initialize(W.slice(i));
}

// Convenience typedefs.

using XavierInitialization = GlorotInitializationType<true>;

using GlorotInitialization = GlorotInitializationType<false>;
// Uses normal distribution
} // namespace ann
} // namespace mlpack

#endif