akmbiclust {akmbiclust}

Alternating k-means biclustering

Description

This function uses the alternating k-means biclustering algorithm to extract the k biclusters in the matrix X. See the paper "Biclustering with Alternating K-Means" for more details.

Usage

akmbiclust(X, k, lambda = 0, nstart = 1)

Arguments

X

Data matrix.

k

The number of biclusters.

lambda

Regularization parameter. Default is 0.

nstart

The number of random initializations. Default is 1.

Value

A list containing three objects:

row_labels

The bicluster labels of every row.

col_labels

The bicluster labels of every column.

loss

The loss of the produced biclusters.

Author(s)

Nicolas Fraiman and Zichao Li

References

N. Fraiman and Z. Li (2020). Biclustering with Alternating K-Means. arXiv preprint arXiv:2009.04550.

Examples

# we create a 100 by 100 matrix X which has an underlying 2 by 2 block structure.
# The entries in the two 50 by 50 blocks on the top left and bottom right follow
# i.i.d. normal with mean 0 and variance 4. The entries in the two 50 by 50 blocks
# on the top right and bottom left follow i.i.d. normal with mean 0 and variance 1.

X <- matrix(rnorm(10000, 0, 1), 100, 100)
X[1:50, 1:50] <- matrix(rnorm(2500, 0, 2), 50, 50)
X[51:100, 51:100] <- matrix(rnorm(2500, 0, 2), 50, 50)

# Alternating k-means biclustering
# Result: perfect
result <- akmbiclust(X, 2, lambda = 0, nstart = 100)
result$row_labels
result$col_labels

# Separate k-means clustering on the rows and columns
# Result: random
kmeans(X, 2)$cluster
kmeans(t(X), 2)$cluster