Ensemble Methods for Bicluster Algorithms

Share:

Description

Calculates an ensemble of biclusters from different parameter setting of possible different bicluster algorithms.

Usage

1
2
ensemble(x, confs, rep = 1, maxNum = 5, similar = jaccard2, thr = 0.8, simthr =0.7,
  subs = c(1, 1), bootstrap = FALSE, support = 0, combine=firstcome, ...)

Arguments

x

Data Matrix

confs

Matrix containing parameter sets

rep

Number of repetitions for each parameter set

maxNum

Maximum number of biclusters taken from each run

similar

Function to produce a similarity matrix of bicluster

thr

Threshold for similarity

simthr

Proportion of row column combinations in bicluster

subs

Vector of proportion of rows and columns for subsampling. Default c(1,1) means no subsampling.

bootstrap

Should bootstrap sampling be used (logical: replace=bootstrap).

support

Wich proportion of the runs must contain the bicluster to have enough support to report it (between 0 and 1).

combine

Function to combine the single bicluster only firstcome and hcl for hierarchical clustering are possible at the moment.

...

Arguments past to the combine function.

Details

Two different kinds (or both combined) of ensebmbling is possible. Ensemble of repeated runs or ensemble of runs on subsamples.

Value

Return an object of class Biclust

Author(s)

Sebastian Kaiser sebastian.kaiser@stat.uni-muenchen.de

See Also

Biclust-class, plaid.grid, bimax.grid

Examples

1
2
3
4
5
6
data(BicatYeast)
ensemble.plaid <- ensemble(BicatYeast,plaid.grid()[1:5],rep=1,maxNum=2, thr=0.5, subs = c(1,1))
ensemble.plaid
x <- binarize(BicatYeast)
ensemble.bimax <- ensemble(x,bimax.grid(),rep=10,maxNum=2,thr=0.5, subs = c(0.8,0.8))
ensemble.bimax