GPUMLib  0.2.2
GPU Machine Learning Library
Public Member Functions | Protected Member Functions | Protected Attributes | List of all members
NMF_MultiplicativeEuclidianDistance Class Reference

Represents a Non-Negative Matrix Factorization (NMF) algorithm that uses multiplicative update rules and the Euclidean distance metric. More...

#include <NMFmultiplicativeEuclidian.h>

Inheritance diagram for NMF_MultiplicativeEuclidianDistance:
NMF

Public Member Functions

 NMF_MultiplicativeEuclidianDistance (HostMatrix< cudafloat > &v, int r)
 
 NMF_MultiplicativeEuclidianDistance (HostMatrix< cudafloat > &v, HostMatrix< cudafloat > &w, HostMatrix< cudafloat > &h)
 
void DoIteration (bool updateW=true)
 
HostMatrix< cudafloatGetW ()
 
HostMatrix< cudafloatGetH ()
 
HostMatrix< cudafloatGetWH ()
 
cudafloat QualityImprovement () const
 Gets the quality improvement caused by the last iteration.
 

Protected Member Functions

void DetermineQualityImprovement (bool calculateWH)
 

Protected Attributes

DeviceMatrix< cudafloatV
 
DeviceMatrix< cudafloatW
 
DeviceMatrix< cudafloatH
 
DeviceMatrix< cudafloatWH
 

Detailed Description

Represents a Non-Negative Matrix Factorization (NMF) algorithm that uses multiplicative update rules and the Euclidean distance metric.

Definition at line 33 of file NMFmultiplicativeEuclidian.h.

Constructor & Destructor Documentation

Constructs a Non-negative Matrix Factorization (NMF) algorithm object that uses multiplicative update rules and the Euclidean distance metric. Given a non-negative matrix V (n x m), the NMF algorithm will find non-negative matrix factors W (n x r) and H (r x m) such that V is approximately equal to WH.

Parameters
vn x m matrix (V) containing a set of multivariate n-dimensional data vectors. m is the number of examples in the dataset.
rDetermines the size of matrices W (n x r) and H (r x m).
Attention
Matrix V must be in column-major order

Definition at line 50 of file NMFmultiplicativeEuclidian.h.

Constructs a Non-negative Matrix Factorization (NMF) algorithm object that uses multiplicative update rules and the Euclidean distance metric. Given a non-negative matrix V (n x m), the NMF algorithm will find non-negative matrix factors W (n x r) and H (r x m) such that V is approximately equal to WH.

Parameters
vn x m matrix (V) containing a set of multivariate n-dimensional data vectors. m is the number of examples in the dataset.
wn x r matrix.
hr x m matrix.
Attention
Matrix V must be in column-major order

Definition at line 61 of file NMFmultiplicativeEuclidian.h.

Member Function Documentation

HostMatrix<cudafloat> GetH ( )
inlineinherited

Gets the H matrix

Returns
the H matrix

Definition at line 126 of file BaseNMF.h.

HostMatrix<cudafloat> GetW ( )
inlineinherited

Gets the W matrix

Returns
the W matrix

Definition at line 120 of file BaseNMF.h.

HostMatrix<cudafloat> GetWH ( )
inlineinherited

Gets the approximation, given by WH, to the matrix V

Returns
the approximation, given by WH, to the matrix V

Definition at line 132 of file BaseNMF.h.


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