mlpack  2.2.5
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Pages
quic_svd_method.hpp
Go to the documentation of this file.
1 
14 #ifndef MLPACK_METHODS_PCA_DECOMPOSITION_POLICIES_QUIC_SVD_METHOD_HPP
15 #define MLPACK_METHODS_PCA_DECOMPOSITION_POLICIES_QUIC_SVD_METHOD_HPP
16 
17 #include <mlpack/prereqs.hpp>
19 
20 namespace mlpack {
21 namespace pca {
22 
27 {
28  public:
29 
36  QUICSVDPolicy(const double epsilon = 0.03, const double delta = 0.1) :
37  epsilon(epsilon),
38  delta(delta)
39  {
40  /* Nothing to do here */
41  }
42 
54  void Apply(const arma::mat& data,
55  const arma::mat& centeredData,
56  arma::mat& transformedData,
57  arma::vec& eigVal,
58  arma::mat& eigvec,
59  const size_t /* rank */)
60  {
61  // This matrix will store the right singular values; we do not need them.
62  arma::mat v, sigma;
63 
64  // Do singular value decomposition using the QUIC-SVD algorithm.
65  svd::QUIC_SVD quicsvd(centeredData, eigvec, v, sigma, epsilon, delta);
66 
67  // Now we must square the singular values to get the eigenvalues.
68  // In addition we must divide by the number of points, because the
69  // covariance matrix is X * X' / (N - 1).
70  eigVal = arma::pow(arma::diagvec(sigma), 2) / (data.n_cols - 1);
71 
72  // Project the samples to the principals.
73  transformedData = arma::trans(eigvec) * centeredData;
74  }
75 
77  double Epsilon() const { return epsilon; }
79  double& Epsilon() { return epsilon; }
80 
82  double Delta() const { return delta; }
84  double& Delta() { return delta; }
85 
86  private:
88  double epsilon;
89 
91  double delta;
92 };
93 
94 } // namespace pca
95 } // namespace mlpack
96 
97 #endif
double & Epsilon()
Modify the error tolerance fraction for calculated subspace.
void Apply(const arma::mat &data, const arma::mat &centeredData, arma::mat &transformedData, arma::vec &eigVal, arma::mat &eigvec, const size_t)
Apply Principal Component Analysis to the provided data set using the QUIC-SVD method.
double Epsilon() const
Get the error tolerance fraction for calculated subspace.
The core includes that mlpack expects; standard C++ includes and Armadillo.
QUICSVDPolicy(const double epsilon=0.03, const double delta=0.1)
Use QUIC-SVD method to perform the principal components analysis (PCA).
double Delta() const
Get the cumulative probability for Monte Carlo error lower bound.
double & Delta()
Modify the cumulative probability for Monte Carlo error lower bound.
QUIC-SVD is a matrix factorization technique, which operates in a subspace such that A&#39;s approximatio...
Definition: quic_svd.hpp:53
Implementation of the QUIC-SVD policy.