svd_incomplete_incremental_learning.hpp

Go to the documentation of this file.

#ifndef MLPACK_METHODS_AMF_SVD_INCOMPLETE_INCREMENTAL_LEARNING_HPP
#define MLPACK_METHODS_AMF_SVD_INCOMPLETE_INCREMENTAL_LEARNING_HPP

namespace mlpack
{
namespace amf
{

class SVDIncompleteIncrementalLearning
{
 public:
  SVDIncompleteIncrementalLearning(double u = 0.001,
                                   double kw = 0,
                                   double kh = 0)
          : u(u), kw(kw), kh(kh), currentUserIndex(0)
  {
    // Nothing to do.
  }

  template<typename MatType>
  void Initialize(const MatType& /* dataset */, const size_t /* rank */)
  {
    // Set the current user to 0.
    currentUserIndex = 0;
  }

  template<typename MatType>
  inline void WUpdate(const MatType& V,
                      arma::mat& W,
                      const arma::mat& H)
  {
    arma::mat deltaW;
    deltaW.zeros(V.n_rows, W.n_cols);

    // Iterate through all the rating by this user to update corresponding item
    // feature feature vector.
    for (size_t i = 0; i < V.n_rows; ++i)
    {
      const double val = V(i, currentUserIndex);
      // Update only if the rating is non-zero.
      if (val != 0)
      {
        deltaW.row(i) += (val - arma::dot(W.row(i), H.col(currentUserIndex))) *
            H.col(currentUserIndex).t();
      }
      // Add regularization.
      if (kw != 0)
        deltaW.row(i) -= kw * W.row(i);
    }

    W += u * deltaW;
  }

  template<typename MatType>
  inline void HUpdate(const MatType& V,
                      const arma::mat& W,
                      arma::mat& H)
  {
    arma::vec deltaH;
    deltaH.zeros(H.n_rows);

    // Iterate through all the rating by this user to update corresponding item
    // feature feature vector.
    for (size_t i = 0; i < V.n_rows; ++i)
    {
      const double val = V(i, currentUserIndex);
      // Update only if the rating is non-zero.
      if (val != 0)
      {
        deltaH += (val - arma::dot(W.row(i), H.col(currentUserIndex))) *
            W.row(i).t();
      }
    }
    // Add regularization.
    if (kh != 0)
      deltaH -= kh * H.col(currentUserIndex);

    // Update H matrix and move on to the next user.
    H.col(currentUserIndex++) += u * deltaH;
    currentUserIndex = currentUserIndex % V.n_cols;
  }

 private:
  double u;
  double kw;
  double kh;

  size_t currentUserIndex;
};


template<>
inline void SVDIncompleteIncrementalLearning::WUpdate<arma::sp_mat>(
    const arma::sp_mat& V, arma::mat& W, const arma::mat& H)
{
  arma::mat deltaW(V.n_rows, W.n_cols);
  deltaW.zeros();
  for (arma::sp_mat::const_iterator it = V.begin_col(currentUserIndex);
      it != V.end_col(currentUserIndex); ++it)
  {
    double val = *it;
    size_t i = it.row();
    deltaW.row(i) += (val - arma::dot(W.row(i), H.col(currentUserIndex))) *
        arma::trans(H.col(currentUserIndex));
    if (kw != 0) deltaW.row(i) -= kw * W.row(i);
  }

  W += u*deltaW;
}

template<>
inline void SVDIncompleteIncrementalLearning::HUpdate<arma::sp_mat>(
    const arma::sp_mat& V, const arma::mat& W, arma::mat& H)
{
  arma::mat deltaH(H.n_rows, 1);
  deltaH.zeros();

  for (arma::sp_mat::const_iterator it = V.begin_col(currentUserIndex);
      it != V.end_col(currentUserIndex); ++it)
  {
    double val = *it;
    size_t i = it.row();
    if ((val = V(i, currentUserIndex)) != 0)
    {
      deltaH += (val - arma::dot(W.row(i), H.col(currentUserIndex))) *
          arma::trans(W.row(i));
    }
  }
  if (kh != 0) deltaH -= kh * H.col(currentUserIndex);

  H.col(currentUserIndex++) += u * deltaH;
  currentUserIndex = currentUserIndex % V.n_cols;
}

} // namespace amf
} // namespace mlpack

#endif