lmkkmeans_missingData {klic}

Localised multiple kernel k-means

Description

Perform the training step of the localised multiple kernel k-means.

Usage

lmkkmeans_missingData(Km, parameters, missing = NULL, verbose = FALSE)

Arguments

Km

Array of size N X N X M containing M different N x N kernel matrices.

parameters

A list of parameters containing the desired number of clusters, cluster_count, and the number of iterations of the algorithm to be run, iteration_count.

missing

Matrix of size N X M containing missingness indicators, i.e. missing[i,j] = 1 (or = TRUE) if observation i is missing in dataset j, missing[i,j] = 0 (or = FALSE).

verbose

Boolean flag. If TRUE, at each iteration the iteration number is printed. Defaults to FALSE.

Value

This function returns a list containing:

clustering

the cluster labels for each element (i.e. row/column) of the kernel matrix.

objective

the value of the objective function for the given clustering.

parameters

same parameters as in the input.

Theta

N x M matrix of weights, each row corresponds to an observation and each column to one of the kernels.

Author(s)

Mehmet Gonen, Alessandra Cabassi

References

Gonen, M. and Margolin, A.A., 2014. Localized data fusion for kernel k-means clustering with application to cancer biology. In Advances in Neural Information Processing Systems (pp. 1305-1313).

Examples

if(requireNamespace("Rmosek", quietly = TRUE) &&
(!is.null(utils::packageDescription("Rmosek")$Configured.MSK_VERSION))){

# Intialise 100 x 100 x 3 array containing M kernel matrices
# representing three different types of similarities between 100 data points
km <- array(NA, c(100, 100, 3))
# Load kernel matrices
km[,,1] <- as.matrix(read.csv(system.file('extdata',
'kernel_matrix1.csv', package = 'klic'), row.names = 1))
km[,,2] <- as.matrix(read.csv(system.file('extdata',
'kernel_matrix2.csv', package = 'klic'), row.names = 1))
km[,,3] <- as.matrix(read.csv(system.file('extdata',
'kernel_matrix3.csv', package = 'klic'), row.names = 1))
# Introduce some missing data
km[76:80, , 1] <- NA
km[, 76:80, 1] <- NA

# Define missingness indicators
missing <- matrix(FALSE, 100, 3)
missing[76:80,1] <- TRUE

# Initalize the parameters of the algorithm
parameters <- list()
# Set the number of clusters
parameters$cluster_count <- 4
# Set the number of iterations
parameters$iteration_count <- 10

# Perform training
state <- lmkkmeans_missingData(km, parameters, missing)

# Display the clustering
print(state$clustering)
# Display the kernel weights
print(state$Theta)
}