seeding: finding initial cluster centers

In evanwang1990/ClusterTools: Tools for Cluster Analysis

Description

A fast seeding algorithm proposing to approximate k-means++ using Markov chain Monte Carlo (MCMC).

Usage

seeding(X, num_seeds, m, threads)

Arguments

X	a numeric matrix or data frame
num_seeds	number of seeds
m	an integer, length of Markove chains, which is used to sample centers, 20 by default
threads	an integer, number of threads to speed up computing

Examples

data(iris)
iris <- iris[sample(1:nrow(iris), 5000, replace = T),]
X <- as.matrix(iris[,1:4])
seeds <- seeding(X, 3, 20, 2)
clus1 <- kmeans(X, X[seeds,]); table(clus1$cluster, iris[,5])
clus2 <- kmeans(X, 3); table(clus2$cluster, iris[,5])

evanwang1990/ClusterTools documentation built on May 16, 2019, 9:37 a.m.