kebabs (version 1.6.2)

linearKernel: Linear Kernel

Description

Create a dense or sparse kernel matrix from an explicit representation

Usage

linearKernel(x, y = NULL, selx = integer(0), sely = integer(0),
  sparse = FALSE, triangular = TRUE, diag = TRUE, lowerLimit = 0)

Arguments

x

a dense or sparse explicit representation. x must be a sparse explicit representation when a sparse kernel matrix should be returned by the function (see parameter sparse).

y

a dense or sparse explicit representation. If x is dense, y must be dense. If x is sparse, y must be sparse.

selx

a numeric or character vector for defining a subset of x. Default=integer(0)

sely

a numeric or character vector for defining a subset of y. Default=integer(0)

sparse

boolean indicating whether returned kernel matrix should be sparse or dense. For value FALSE a dense kernel matrix of class KernelMatrix is returned. If set to TRUE the kernel matrix is returned as sparse matrix of class dgCMatrix. In case of a symmetric matrix either the lower triangular part or the full matrix can be returned. Please note that a sparse kernel matrix currently can not be used for SVM based learning in kebabs. Default=FALSE

triangular

boolean indicating whether just the lower triangular or the full sparse matrix should be returned. This parameter is only relevant for a sparse symmetric kernel matrix. Default=TRUE

diag

boolean indicating whether the diagonal should be included in a sparse triangular matrix. This parameter is only relevant when parameter sparse and triangular are set to TRUE. Default=TRUE

lowerLimit

a numeric value for a similarity threshold. The parameter is relevant for sparse kernel matrices only. If set to a value larger than 0 only similarity values larger than this threshold will be included in the sparse kernel matrix. Default=0

Value

• linearKernel: kernel matrix as class KernelMatrix or sparse kernel matrix of class dgCMatrix dependent on parameter sparse

References

http://www.bioinf.jku.at/software/kebabs J. Palme, S. Hochreiter, and U. Bodenhofer (2015) KeBABS: an R package for kernel-based analysis of biological sequences. Bioinformatics, 31(15):2574-2576, 2015. DOI: http://dx.doi.org/10.1093/bioinformatics/btv176{10.1093/bioinformatics/btv176}.

Examples

## load sequence data and change sample names
data(TFBS)
names(enhancerFB) <- paste("S", 1:length(enhancerFB), sep="_")

## create the kernel object for dimers with normalization
speck <- spectrumKernel(k=5)

## generate sparse explicit representation
ers <- getExRep(enhancerFB, speck)

## compute dense kernel matrix (as currently used in SVM based learning)
km <- linearKernel(ers)
km[1:5, 1:5]

## compute sparse kernel matrix
## because it is symmetric just the lower diagonal
## is computed to save storage
km <- linearKernel(ers, sparse=TRUE)
km[1:5, 1:5]

## compute full sparse kernel matrix
km <- linearKernel(ers, sparse=TRUE, triangular=FALSE)
km[1:5, 1:5]

## compute triangular sparse kernel matrix without diagonal
km <- linearKernel(ers, sparse=TRUE, triangular=TRUE, diag=FALSE)
km[1:5, 1:5]

## plot histogram of similarity values
hist(as(km, "numeric"), breaks=30)

## compute sparse kernel matrix with similarities above 0.5 only
km <- linearKernel(ers, sparse=TRUE, lowerLimit=0.5)
km[1:5, 1:5]