mlpack 3.4.2
regularized_svd_method.hpp
Go to the documentation of this file.
1
14#ifndef MLPACK_METHODS_CF_DECOMPOSITION_POLICIES_REGULARIZED_SVD_METHOD_HPP
15#define MLPACK_METHODS_CF_DECOMPOSITION_POLICIES_REGULARIZED_SVD_METHOD_HPP
16
17#include <mlpack/prereqs.hpp>
19
20namespace mlpack {
21namespace cf {
22
42{
43 public:
50 RegSVDPolicy(const size_t maxIterations = 10) :
51 maxIterations(maxIterations)
52 {
53 /* Nothing to do here */
54 }
55
68 void Apply(const arma::mat& data,
69 const arma::sp_mat& /* cleanedData */,
70 const size_t rank,
71 const size_t maxIterations,
72 const double /* minResidue */,
73 const bool /* mit */)
74 {
75 // Do singular value decomposition using the regularized SVD algorithm.
76 svd::RegularizedSVD<> regsvd(maxIterations);
77 regsvd.Apply(data, rank, w, h);
78 }
79
86 double GetRating(const size_t user, const size_t item) const
87 {
88 double rating = arma::as_scalar(w.row(item) * h.col(user));
89 return rating;
90 }
91
98 void GetRatingOfUser(const size_t user, arma::vec& rating) const
99 {
100 rating = w * h.col(user);
101 }
102
115 template<typename NeighborSearchPolicy>
116 void GetNeighborhood(const arma::Col<size_t>& users,
117 const size_t numUsersForSimilarity,
118 arma::Mat<size_t>& neighborhood,
119 arma::mat& similarities) const
120 {
121 // We want to avoid calculating the full rating matrix, so we will do
122 // nearest neighbor search only on the H matrix, using the observation that
123 // if the rating matrix X = W*H, then d(X.col(i), X.col(j)) = d(W H.col(i),
124 // W H.col(j)). This can be seen as nearest neighbor search on the H
125 // matrix with the Mahalanobis distance where M^{-1} = W^T W. So, we'll
126 // decompose M^{-1} = L L^T (the Cholesky decomposition), and then multiply
127 // H by L^T. Then we can perform nearest neighbor search.
128 arma::mat l = arma::chol(w.t() * w);
129 arma::mat stretchedH = l * h; // Due to the Armadillo API, l is L^T.
130
131 // Temporarily store feature vector of queried users.
132 arma::mat query(stretchedH.n_rows, users.n_elem);
133 // Select feature vectors of queried users.
134 for (size_t i = 0; i < users.n_elem; ++i)
135 query.col(i) = stretchedH.col(users(i));
136
137 NeighborSearchPolicy neighborSearch(stretchedH);
138 neighborSearch.Search(
139 query, numUsersForSimilarity, neighborhood, similarities);
140 }
141
143 const arma::mat& W() const { return w; }
145 const arma::mat& H() const { return h; }
146
148 size_t MaxIterations() const { return maxIterations; }
150 size_t& MaxIterations() { return maxIterations; }
151
155 template<typename Archive>
156 void serialize(Archive& ar, const unsigned int /* version */)
157 {
158 ar & BOOST_SERIALIZATION_NVP(w);
159 ar & BOOST_SERIALIZATION_NVP(h);
160 }
161
162 private:
164 size_t maxIterations;
166 arma::mat w;
168 arma::mat h;
169};
170
171} // namespace cf
172} // namespace mlpack
173
174#endif
Implementation of the Regularized SVD policy to act as a wrapper when accessing Regularized SVD from ...
double GetRating(const size_t user, const size_t item) const
Return predicted rating given user ID and item ID.
void GetNeighborhood(const arma::Col< size_t > &users, const size_t numUsersForSimilarity, arma::Mat< size_t > &neighborhood, arma::mat &similarities) const
Get the neighborhood and corresponding similarities for a set of users.
void Apply(const arma::mat &data, const arma::sp_mat &, const size_t rank, const size_t maxIterations, const double, const bool)
Apply Collaborative Filtering to the provided data set using the regularized SVD.
size_t MaxIterations() const
Get the number of iterations.
size_t & MaxIterations()
Modify the number of iterations.
const arma::mat & W() const
Get the Item Matrix.
RegSVDPolicy(const size_t maxIterations=10)
Use regularized SVD method to perform collaborative filtering.
const arma::mat & H() const
Get the User Matrix.
void GetRatingOfUser(const size_t user, arma::vec &rating) const
Get predicted ratings for a user.
void serialize(Archive &ar, const unsigned int)
Serialization.
Regularized SVD is a matrix factorization technique that seeks to reduce the error on the training se...
void Apply(const arma::mat &data, const size_t rank, arma::mat &u, arma::mat &v)
Obtains the user and item matrices using the provided data and rank.
Linear algebra utility functions, generally performed on matrices or vectors.
Definition: cv.hpp:1
The core includes that mlpack expects; standard C++ includes and Armadillo.