SYNOPSIS

Public Member Functions

NystroemMethod (const arma::mat &data, KernelType &kernel, const size_t rank)
Create the NystroemMethod object.
void Apply (arma::mat &output)
Apply the low-rank factorization to obtain an output matrix G such that K' = G * G^T.
void GetKernelMatrix (const arma::mat *data, arma::mat &miniKernel, arma::mat &semiKernel)
Construct the kernel matrix with matrix that contains the selected points.
void GetKernelMatrix (const arma::Col< size_t > &selectedPoints, arma::mat &miniKernel, arma::mat &semiKernel)
Construct the kernel matrix with the selected points.

Private Attributes

const arma::mat & data
The reference dataset.
KernelType & kernel
The locally stored kernel, if it is necessary.
const size_t rank
Rank used for matrix approximation.

Detailed Description

template<typename KernelType, typename PointSelectionPolicy = KMeansSelection<>>class mlpack::kernel::NystroemMethod< KernelType, PointSelectionPolicy >

Definition at line 38 of file nystroem_method.hpp.

Constructor & Destructor Documentation

template<typename KernelType, typename PointSelectionPolicy = KMeansSelection<>> mlpack::kernel::NystroemMethod< KernelType, PointSelectionPolicy >::NystroemMethod (const arma::mat &data, KernelType &kernel, const size_trank)

Create the NystroemMethod object. The constructor here does not really do anything.

Parameters:

data Data matrix.
kernel Kernel to be used for computation.
rank Rank to be used for matrix approximation.

Member Function Documentation

template<typename KernelType, typename PointSelectionPolicy = KMeansSelection<>> void mlpack::kernel::NystroemMethod< KernelType, PointSelectionPolicy >::Apply (arma::mat &output)

Apply the low-rank factorization to obtain an output matrix G such that K' = G * G^T.

Parameters:

output Matrix to store kernel approximation into.

Referenced by mlpack::kpca::NystroemKernelRule< KernelType, PointSelectionPolicy >::ApplyKernelMatrix().

template<typename KernelType, typename PointSelectionPolicy = KMeansSelection<>> void mlpack::kernel::NystroemMethod< KernelType, PointSelectionPolicy >::GetKernelMatrix (const arma::mat *data, arma::mat &miniKernel, arma::mat &semiKernel)

Construct the kernel matrix with matrix that contains the selected points.

Parameters:

data Data matrix pointer.
miniKernel to store the constructed mini-kernel matrix in.
miniKernel to store the constructed semi-kernel matrix in.

template<typename KernelType, typename PointSelectionPolicy = KMeansSelection<>> void mlpack::kernel::NystroemMethod< KernelType, PointSelectionPolicy >::GetKernelMatrix (const arma::Col< size_t > &selectedPoints, arma::mat &miniKernel, arma::mat &semiKernel)

Construct the kernel matrix with the selected points.

Parameters:

points Indices of selected points.
miniKernel to store the constructed mini-kernel matrix in.
miniKernel to store the constructed semi-kernel matrix in.

Member Data Documentation

template<typename KernelType, typename PointSelectionPolicy = KMeansSelection<>> const arma::mat& mlpack::kernel::NystroemMethod< KernelType, PointSelectionPolicy >::data [private]

The reference dataset.

Definition at line 83 of file nystroem_method.hpp.

template<typename KernelType, typename PointSelectionPolicy = KMeansSelection<>> KernelType& mlpack::kernel::NystroemMethod< KernelType, PointSelectionPolicy >::kernel [private]

The locally stored kernel, if it is necessary.

Definition at line 85 of file nystroem_method.hpp.

template<typename KernelType, typename PointSelectionPolicy = KMeansSelection<>> const size_t mlpack::kernel::NystroemMethod< KernelType, PointSelectionPolicy >::rank [private]

Rank used for matrix approximation.

Definition at line 87 of file nystroem_method.hpp.

Author

Generated automatically by Doxygen for MLPACK from the source code.