max_pooling.hpp

Go to the documentation of this file.

#ifndef MLPACK_METHODS_ANN_LAYER_MAX_POOLING_HPP
#define MLPACK_METHODS_ANN_LAYER_MAX_POOLING_HPP

#include <mlpack/prereqs.hpp>

namespace mlpack {
namespace ann  {

/*
 * The max pooling rule for convolution neural networks. Take the maximum value
 * within the receptive block.
 */
class MaxPoolingRule
{
 public:
  /*
   * Return the maximum value within the receptive block.
   *
   * @param input Input used to perform the pooling operation.
   */
  template<typename MatType>
  size_t Pooling(const MatType& input)
  {
    return arma::as_scalar(arma::find(input.max() == input, 1));
  }
};

template <
    typename InputDataType = arma::mat,
    typename OutputDataType = arma::mat
>
class MaxPooling
{
 public:
  MaxPooling();

  MaxPooling(const size_t kW,
             const size_t kH,
             const size_t dW = 1,
             const size_t dH = 1,
             const bool floor = true);

  template<typename eT>
  void Forward(const arma::Mat<eT>&& input, arma::Mat<eT>&& output);

  template<typename eT>
  void Backward(const arma::Mat<eT>&& /* input */,
                arma::Mat<eT>&& gy,
                arma::Mat<eT>&& g);

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

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

  size_t const& InputWidth() const { return inputWidth; }
  size_t& InputWidth() { return inputWidth; }

  size_t const& InputHeight() const { return inputHeight; }
  size_t& InputHeight() { return inputHeight; }

  size_t const& OutputWidth() const { return outputWidth; }
  size_t& OutputWidth() { return outputWidth; }

  size_t const& OutputHeight() const { return outputHeight; }
  size_t& OutputHeight() { return outputHeight; }

  bool Deterministic() const { return deterministic; }
  bool& Deterministic() { return deterministic; }

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

 private:
  template<typename eT>
  void PoolingOperation(const arma::Mat<eT>& input,
                        arma::Mat<eT>& output,
                        arma::Mat<eT>& poolingIndices)
  {
    for (size_t j = 0, colidx = 0; j < output.n_cols; ++j, colidx += dW)
    {
      for (size_t i = 0, rowidx = 0; i < output.n_rows; ++i, rowidx += dH)
      {
        arma::mat subInput = input(arma::span(rowidx, rowidx + kW - 1 - offset),
            arma::span(colidx, colidx + kH - 1 - offset));

        const size_t idx = pooling.Pooling(subInput);
        output(i, j) = subInput(idx);

        if (!deterministic)
        {
          arma::Mat<size_t> subIndices = indices(arma::span(rowidx,
              rowidx + kW - 1 - offset),
              arma::span(colidx, colidx + kH - 1 - offset));

          poolingIndices(i, j) = subIndices(idx);
        }
      }
    }
  }

  template<typename eT>
  void Unpooling(const arma::Mat<eT>& error,
                 arma::Mat<eT>& output,
                 arma::Mat<eT>& poolingIndices)
  {
    for (size_t i = 0; i < poolingIndices.n_elem; ++i)
    {
      output(poolingIndices(i)) += error(i);
    }
  }

  size_t kW;

  size_t kH;

  size_t dW;

  size_t dH;

  bool floor;

  size_t inSize;

  size_t outSize;

  bool reset;

  size_t inputWidth;

  size_t inputHeight;

  size_t outputWidth;

  size_t outputHeight;

  bool deterministic;

  size_t offset;

  size_t batchSize;

  arma::cube outputTemp;

  arma::cube inputTemp;

  arma::cube gTemp;

  MaxPoolingRule pooling;

  OutputDataType delta;

  OutputDataType gradient;

  OutputDataType outputParameter;

  arma::Mat<size_t> indices;

  arma::Col<size_t> indicesCol;

  std::vector<arma::cube> poolingIndices;
}; // class MaxPooling

} // namespace ann
} // namespace mlpack

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

#endif