mlpack  master
KernelPCA< KernelType, KernelRule > Class Template Reference

This class performs kernel principal components analysis (Kernel PCA), for a given kernel. More...

Public Member Functions

 KernelPCA (const KernelType kernel=KernelType(), const bool centerTransformedData=false)
 Construct the KernelPCA object, optionally passing a kernel. More...

 
void Apply (const arma::mat &data, arma::mat &transformedData, arma::vec &eigval, arma::mat &eigvec, const size_t newDimension)
 Apply Kernel Principal Components Analysis to the provided data set. More...

 
void Apply (const arma::mat &data, arma::mat &transformedData, arma::vec &eigval, arma::mat &eigvec)
 Apply Kernel Principal Components Analysis to the provided data set. More...

 
void Apply (const arma::mat &data, arma::mat &transformedData, arma::vec &eigval)
 Apply Kernel Principal Component Analysis to the provided data set. More...

 
void Apply (arma::mat &data, const size_t newDimension)
 Apply dimensionality reduction using Kernel Principal Component Analysis to the provided data set. More...

 
bool CenterTransformedData () const
 Return whether or not the transformed data is centered. More...

 
bool & CenterTransformedData ()
 Return whether or not the transformed data is centered. More...

 
const KernelType & Kernel () const
 Get the kernel. More...

 
KernelType & Kernel ()
 Modify the kernel. More...

 

Detailed Description


template<typename KernelType, typename KernelRule = NaiveKernelRule<KernelType>>
class mlpack::kpca::KernelPCA< KernelType, KernelRule >

This class performs kernel principal components analysis (Kernel PCA), for a given kernel.

This is a standard machine learning technique and is well-documented on the Internet and in standard texts. It is often used as a dimensionality reduction technique, and can also be useful in mapping linearly inseparable classes of points to different spaces where they are linearly separable.

The performance of the method is highly dependent on the kernel choice. There are numerous available kernels in the mlpack::kernel namespace (see files in mlpack/core/kernels/) and it is easy to write your own; see other implementations for examples.

Definition at line 40 of file kernel_pca.hpp.

Constructor & Destructor Documentation

◆ KernelPCA()

KernelPCA ( const KernelType  kernel = KernelType(),
const bool  centerTransformedData = false 
)

Construct the KernelPCA object, optionally passing a kernel.

Optionally, the transformed data can be centered about the origin; to do this, pass 'true' for centerTransformedData. This will take slightly longer (but not much).

Parameters
kernelKernel to be used for computation.
centerTransformedDataCenter transformed data.

Member Function Documentation

◆ Apply() [1/4]

void Apply ( const arma::mat &  data,
arma::mat &  transformedData,
arma::vec &  eigval,
arma::mat &  eigvec,
const size_t  newDimension 
)

Apply Kernel Principal Components Analysis to the provided data set.

Parameters
dataData matrix.
transformedDataMatrix to output results into.
eigvalKPCA eigenvalues will be written to this vector.
eigvecKPCA eigenvectors will be written to this matrix.
newDimensionNew dimension for the dataset.

◆ Apply() [2/4]

void Apply ( const arma::mat &  data,
arma::mat &  transformedData,
arma::vec &  eigval,
arma::mat &  eigvec 
)

Apply Kernel Principal Components Analysis to the provided data set.

Parameters
dataData matrix.
transformedDataMatrix to output results into.
eigvalKPCA eigenvalues will be written to this vector.
eigvecKPCA eigenvectors will be written to this matrix.

◆ Apply() [3/4]

void Apply ( const arma::mat &  data,
arma::mat &  transformedData,
arma::vec &  eigval 
)

Apply Kernel Principal Component Analysis to the provided data set.

Parameters
dataData matrix.
transformedDataMatrix to output results into.
eigvalKPCA eigenvalues will be written to this vector.

◆ Apply() [4/4]

void Apply ( arma::mat &  data,
const size_t  newDimension 
)

Apply dimensionality reduction using Kernel Principal Component Analysis to the provided data set.

The data matrix will be modified in-place. Note that the dimension can be larger than the existing dimension because KPCA works on the kernel matrix, not the covariance matrix. This means the new dimension can be as large as the number of points (columns) in the dataset. Note that if you specify newDimension to be larger than the current dimension of the data (the number of rows), then it's not really "dimensionality reduction"...

Parameters
dataData matrix.
newDimensionNew dimension for the dataset.

◆ CenterTransformedData() [1/2]

bool CenterTransformedData ( ) const
inline

Return whether or not the transformed data is centered.

Definition at line 115 of file kernel_pca.hpp.

◆ CenterTransformedData() [2/2]

bool& CenterTransformedData ( )
inline

Return whether or not the transformed data is centered.

Definition at line 117 of file kernel_pca.hpp.

◆ Kernel() [1/2]

const KernelType& Kernel ( ) const
inline

Get the kernel.

Definition at line 110 of file kernel_pca.hpp.

◆ Kernel() [2/2]

KernelType& Kernel ( )
inline

Modify the kernel.

Definition at line 112 of file kernel_pca.hpp.


The documentation for this class was generated from the following file: