Sketch: Sketch

In chickn: 'Compressive' Hierarchical Kernel Clustering Toolbox

Description

The data sketch computation.

Usage

Sketch(Data, W, ind.col = 1:ncol(Data), ncores = 1, parallel = FALSE)

Arguments

Data	A Filebacked Big Matrix n x N. Data signals are stored in the matrix columns.
W	A frequency matrix m x n. The frequency vectors are stored in the matrix rows.
ind.col	Column indeces for which the data sketch is computed. By default all matrix columns.
ncores	Number of used cores. By default 1. If parallel = FALSE, ncores defines a number of data splits on which the sketch is computed separatelly.
parallel	logical parameter that indicates whether computations are performed on several cores in parallel or not.

Details

The sketch of the given data collection x_1, …, x_N is a vector of the length 2m. First m components of the data sketch vector correspond to its real part, i.e. \frac{1}{N} ∑_{i=1}^N \cos(W x_i). Last m components are its imaginary part, i.e. \frac{1}{N} ∑_{i=1}^N \sin(W x_i).

Value

The data sketch vector.

References

\insertRef

DBLP:journals/corr/KerivenBGP16chickn.

Examples

X = matrix(rnorm(1000), ncol=100, nrow = 10)
X_FBM = bigstatsr::FBM(init = X, ncol=100, nrow = 10)
W = GenerateFrequencies(Data = X_FBM, m = 20, N0 = 100, TypeDist = "AR")$W
SK1 = Sketch(X_FBM, W)
SK2 = Sketch(X_FBM, W, parallel = TRUE, ncores = 2)
all.equal(SK1, SK2)

chickn documentation built on Jan. 13, 2021, 10:53 p.m.