rCUR (version 1.3)

rCUR-package: CUR matrix decomposition package

Description

In modern data mining tasks the user has often matrices with huge number of rows and/or columns as the base of analysis. One way of the analysis of this type of huge datasets is to reduct their dimensions. Principal component analysis (PCA) is a widely used tool for such data analysis. PCA produced singular vectors are mathematical abstractions and hardly interpretable on the field from which the data are drawn. Mahoney & Drineas (2009) proposed a method the CUR matrix decomposition what decreases the dimensions as well, but the resulting matrices are interpretable on the application area. This package contains functions and objects to help doing CUR matrix decomposition.

Details

Package:

rCUR

Type:

Package

Version:

1.0

Date:

2011-03-31

License:

GPL (>=2)

LazyLoad:

yes

References

Mahoney M. W. and Drineas P. (2009) CUR matrix decompositions for improved data analysis. PNAS, 106(3):697-702

Andras Bodor, Istvan Csabai, Michael W Mahoney and Norbert Solymosi rCUR:an R package for CUR matrix decomposition BMC Bioinformatics 2012, 13:103 doi:10.1186/1471-2105-13-103

http://www.cs.rpi.edu/~boutsc/files/AlgorithmCUR.m

Examples

library(rCUR)
library(lattice)

data(STTm)
data(STTa)

n = 27
res = CUR(STTm, 31, n, 4)

plotLeverage(res, C=FALSE, top.n=n, xlab='Gene ID', las=1, top.col='black', 
  top.pch=16, ul.col='black', ul.lty=2, col='grey')

PCA01 = prcomp(STTm, scale=TRUE)

top.idx = topLeverage(res, C=FALSE, top.n=n, sort=FALSE)
top.STTm = STTm[top.idx,]
PCA02 = prcomp(top.STTm, scale=TRUE)

t.top.STTm = t(top.STTm)
x = 1:dim(t.top.STTm)[1]
y = 1:dim(t.top.STTm)[2]

ann.tab = STTa[top.idx,]
genes = as.character(ann.tab$Gene)
genes[genes==''] = as.character(ann.tab[genes=='','Hs'])

op = par(mar=c(3.5, 0, 0, 5) + 0.1)
image(x, y, t.top.STTm, col=rev(gray(0:256/300)), axes=FALSE, xlab='', 
  ylab='')
axis(1, at=c(10,22,31), labels=c('GIST','LEIO','SARC'), tick=FALSE, 
  las=2, cex=0.6, line=-0.8)
axis(4, at=1:length(genes), labels=genes, tick=FALSE, las=1, cex=0.6, 
  line=-0.8)
par(op)

PCAs = rbind(PCA01$rotation[,c(1,2)], PCA02$rotation[,c(1,2)])
group = rownames(PCAs)
pchs = ifelse(group=='GIST', 16, 0)
pchs = ifelse(group=='LEIO', 17, pchs)
pchs = ifelse(group=='SARC', 18, pchs)
PCAs = data.frame(PC1=PCAs[,1], PC2=PCAs[,2], 
  lab=c(rep('all',31),rep('selected',31)), group, pchs)

xyplot(PC2 ~ PC1 | lab, data = PCAs, pch=pchs, cex=1.0, col='black', 
  fill='black', key = list(space = "bottom", text = list(levels(PCAs$group)), 
  points=TRUE, pch=c(16, 17, 18), fill='black', columns=3, just=1))