In jlin-vt/SML: Statistical Machine Learning

knitr::opts_chunk$set(echo = TRUE)
library(SML)
library(ggplot2)

1. Example: Kmeans

Read faithful (2-d) data set and set initial values.

data <- faithful
data <- as.matrix(data) 

K <- 2
centers <- data[sample(nrow(data), K), ]

Run Kmenas:

res <- Kmeans(data, centers, Euclid, 10)
plot(faithful$eruptions, faithful$waiting, 
     xlab = "eruptions", ylab = "waiting",
     col = res$clusters)
points(res$centers, col = 1:2, pch = 8, cex = 2)

2. Example: univariate EM Algorithm

Read faithful (2-d) data set and set initial values.

params1 <- list(mu = c(2, 5), var = c(1, 1), probs = c(.5, .5)) 
params2 <- list(mu = c(50, 90), var = c(1, 15), probs = c(.5, .5))  

X1 <- matrix(faithful[, 1])
X2 <- matrix(faithful[, 2])

Apply algorithm:

test1 <- GaussmixEM(params1, X = X1, tol = 1e-8)
test2 <- GaussmixEM(params2, X = X2, tol = 1e-8)

qplot(x = eruptions, y = waiting, data = faithful)

Make a plot: cluster by eruptions.

ggplot(aes(x = eruptions, y = waiting), data = faithful) +
  geom_point(aes(color = factor(test1$cluster)))  +
  theme(legend.title = element_blank())

Make a plot: cluster by waiting.

ggplot(aes(x = eruptions, y = waiting), data = faithful) +
  geom_point(aes(color = factor(test2$cluster)))  +
  theme(legend.title = element_blank())