conversions: Conversions Between Dense and Sparse Similarity Matrices

In apcluster: Affinity Propagation Clustering

Conversions Between Dense and Sparse Similarity Matrices

Description

Converts a dense similarity matrix into a sparse one or vice versa

Usage

## S4 method for signature 'matrix'
as.SparseSimilarityMatrix(s, lower=-Inf)
## S4 method for signature 'Matrix'
as.SparseSimilarityMatrix(s, lower=-Inf)
## S4 method for signature 'sparseMatrix'
as.SparseSimilarityMatrix(s, lower=-Inf)
## S4 method for signature 'matrix'
as.DenseSimilarityMatrix(s, fill=-Inf)
## S4 method for signature 'Matrix'
as.DenseSimilarityMatrix(s, fill=-Inf)
## S4 method for signature 'sparseMatrix'
as.DenseSimilarityMatrix(s, fill=-Inf)

Arguments

s	a similarity matrix in sparse or dense format (see details below)
lower	cut-off threshold to apply when converting similarity matrices into sparse format. All similarities lower than or equal to lower will be omitted from the result. The default is -Inf), i.e. only -Inf values are removed.
fill	value to fill in for entries that are missing from sparse similarity matrix 's' (defaults to -Inf).

Details

The function as.SparseSimilarityMatrix takes a matrix argument, removes all diagonal elements and all values that are lower than or equal to the cut-off threshold lower and returns a sparse matrix of class dgTMatrix.

If the function as.DenseSimilarityMatrix is called for a sparse matrix (class sparseMatrix or any class derived from this class), a dense matrix is returned, where all values that were missing in the sparse matrix are replaced with fill.

as.DenseSimilarityMatrix can also be called for dense matrix and Matrix objects. In this case, as.DenseSimilarityMatrix assumes that the matrices have three columns that encode for a sparse matrix in the same way as the Matlab implementation of Frey's and Dueck's sparse affinity propagation accepts it: the first column contains 1-based row indices, the second column contains 1-based column indices, and the third column contains the similarity values. The same format is also accepted by as.SparseSimilarityMatrix to convert a sparse similarity matrix of this format into a dgTMatrix object. Note that, for matrices of this format, as.DenseSimilarityMatrix replaces the deprectated function sparseToFull that was used in older versions of the package.

Note that as.SparseSimilarityMatrix and as.DenseSimilarityMatrix are no S4 coercion methods. There are no classes named SparseSimilarityMatrix or DenseSimilarityMatrix.

Value

returns a square similarity matrix in sparse format (class dgTMatrix or in dense format (standard class matrix).

Author(s)

Ulrich Bodenhofer

References

https://github.com/UBod/apcluster

Frey, B. J. and Dueck, D. (2007) Clustering by passing messages between data points. Science 315, 972-976. DOI: \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1126/science.1136800")}.

Bodenhofer, U., Kothmeier, A., and Hochreiter, S. (2011) APCluster: an R package for affinity propagation clustering. Bioinformatics 27, 2463-2464. DOI: \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1093/bioinformatics/btr406")}.

Examples

## create similarity matrix in sparse format according to Frey and Dueck
sp <- matrix(c(1, 2, 0.5, 3, 1, 0.2, 5, 4, -0.2, 3, 4, 1.2), 4, 3, byrow=TRUE)
sp

## perform conversions
as.DenseSimilarityMatrix(sp, fill=0)
as.SparseSimilarityMatrix(sp)

## create dense similarity matrix
cl1 <- cbind(rnorm(20, 0.2, 0.05), rnorm(20, 0.8, 0.06))
cl2 <- cbind(rnorm(20, 0.7, 0.08), rnorm(20, 0.3, 0.05))
x <- rbind(cl1, cl2)

sim <- negDistMat(x, r=2)
ssim <- as.SparseSimilarityMatrix(sim, lower=-0.2)

## run apcluster() on the sparse similarity matrix
apres <- apcluster(ssim, q=0)
apres

apcluster documentation built on May 29, 2024, 2:25 a.m.