mahito-sugiyama

plot2 <- function(..., col_index, draw.axis = TRUE) {
  col_list <- c("#e33518", "#0068a2", "#2b6632", "#9f4a00", "#00a5cb", "#43a347")
  plot(..., pch = 21, col = col_list[col_index], bg = "white", cex = 2.0, axes = FALSE, xlab = "", ylab = "")
  if (draw.axis) axis(1, lwd = 0, lwd.ticks = 1, las = 1, cex.axis = 1.4)
  if (draw.axis) axis(2, lwd = 0, lwd.ticks = 1, las = 1, cex.axis = 1.4)
  if (draw.axis) box(bty = "l")
}
## pdf("file_name.pdf", width = 8, height = 8, useDingbats = FALSE)
## plot2(x)
## dev.off()

mahito-sugiyama

kmeansp2 <- function(x, k, iter.max = 10, nstart = 1, ...) {
  n <- nrow(x) # number of data points
  centers <- numeric(k) # IDs of centers
  distances <- matrix(numeric(n * (k - 1)), ncol = k - 1) # distances[i, j]: The distance between x[i,] and x[centers[j],]
  res.best <- list(tot.withinss = Inf) # the best result among <nstart> iterations
  for (rep in 1:nstart) {
    pr <- rep(1, n) # probability for sampling centers
    for (i in 1:(k - 1)) {
      centers[i] <- sample.int(n, 1, prob = pr) # Pick up the ith center
      distances[, i] <- colSums((t(x) - x[centers[i], ])^2) # Compute (the square of) distances to the center