cobra: Convex Biclustering Algorithm

In cvxbiclustr: Convex Biclustering Algorithm

Description

cobra computes a convex biclustering path via Dykstra-like Proximal Algorithm

Usage

cobra(X, E_row, E_col, w_row, w_col, gamma, max_iter = 100, tol = 0.001)

Arguments

X	The data matrix to be clustered. The rows are the features, and the columns are the samples.
E_row	Edge-incidence matrix for row graph
E_col	Edge-incidence matrix for column graph
w_row	Vector of weights for row graph
w_col	Vector of weights for column graph
gamma	A sequence of regularization parameters for row and column shrinkage
max_iter	Maximum number of iterations
tol	Stopping criterion

Examples

## Create bicluster path
## Example: Lung
X <- lung
X <- X - mean(X)
X <- X/norm(X,'f')

## Create annotation for heatmap
types <- colnames(lung)
ty <- as.numeric(factor(types))
cols <- rainbow(4)
YlGnBu5 <- c('#ffffd9','#c7e9b4','#41b6c4','#225ea8','#081d58')
hmcols <- colorRampPalette(YlGnBu5)(256)

## Construct weights and edge-incidence matrices
phi <- 0.5; k <- 5
wts <- gkn_weights(X,phi=phi,k_row=k,k_col=k)
w_row <- wts$w_row
w_col <- wts$w_col
E_row <- wts$E_row
E_col <- wts$E_col

## Connected Components of Row and Column Graphs
wts$nRowComp
wts$nColComp

#### Initialize path parameters and structures
nGamma <- 5
gammaSeq <- 10**seq(0,3,length.out=nGamma)

## Generate solution path
sol <- cobra_validate(X,E_row,E_col,w_row,w_col,gammaSeq)

ix <- 4
heatmap(sol$U[[ix]],col=hmcols,labRow=NA,labCol=NA,ColSideCol=cols[ty])

Example output

Loading required package: Matrix
Loading required package: igraph

Attaching package: 'igraph'

The following objects are masked from 'package:stats':

    decompose, spectrum

The following object is masked from 'package:base':

    union

[1] 1
[1] 1
[1] "***** Completed gamma = 1 *****"
[1] "***** Completed gamma = 2 *****"
[1] "***** Completed gamma = 3 *****"
[1] "***** Completed gamma = 4 *****"
[1] "***** Completed gamma = 5 *****"

cvxbiclustr documentation built on May 2, 2019, 9:34 a.m.