bdSVD_hdf5: Block SVD decomposition using an incremental algorithm.

In BigDataStatMeth: Statistical Methods and Algorithms for Big Data

Block SVD decomposition using an incremental algorithm.

Description

Block SVD decomposition for hdf5 files using an incremental algorithm.

Usage

bdSVD_hdf5(
  file,
  group = NULL,
  dataset = NULL,
  k = 2L,
  q = 1L,
  bcenter = TRUE,
  bscale = TRUE,
  rankthreshold = 0,
  threads = NULL
)

Arguments

file	a real nxp matrix in hdf5 file
group	group in hdf5 data file where dataset is located
dataset	matrix dataset with data to perform SVD
k	number of local SVDs to concatenate at each level
q	number of levels
bcenter	(optional, defalut = TRUE) . If center is TRUE then centering is done by subtracting the column means (omitting NAs) of x from their corresponding columns, and if center is FALSE, no centering is done.
bscale	(optional, defalut = TRUE) . If scale is TRUE then scaling is done by dividing the (centered) columns of x by their standard deviations if center is TRUE, and the root mean square otherwise. If scale is FALSE, no scaling is done.
rankthreshold	double, threshold used to determine the range of the array. The matrix rank is equal to the number of singular values different from the threshold. By default, threshold = 0 is used to get the matrix rank , but it can be changed to an approximation of 0.
threads	(optional) only used in some operations inside function. If threads is null then threads = maximum number of threads available - 1.

Details

Singular values and left singular vectors of a real nxp matrix

Value

a list of three components with the singular values and left and right singular vectors of the matrix

A List with :

• "u" eigenvectors of AA^t, mxn and column orthogonal matrix

• "v" eigenvectors of A^tA, nxn orthogonal matrix

• "v" singular values, nxn diagonal matrix (non-negative real values)

BigDataStatMeth documentation built on March 30, 2022, 1:07 a.m.