ML-Class Ex 7 - kMeans clustering

The K-means algorithm is a method to automatically cluster similar data examples together.

The intuition behind K-means is an iterative procedure that starts by guessing the initial centroids, and then refines this guess by repeatedly assigning examples to their closest centroids and then recomputing the centroids based on the assignments.

This algorithm was implemented as follows:

?View Code RSPLUS
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
kMeansInitCentroids <- function(X, K) {
    rand.idx <- sample(1:nrow(X), K)
    centroids <- X[rand.idx,]
    return(centroids)
}
 
findClosestCentroids <- function(X, centroids) {
    ## finding closest centroids
 
    # set K
    K <- nrow(centroids)
 
    idx <- sapply(1:nrow(X), function(i) {
        which.min(
                  sapply(1:K, function(j) {
                      sum(
                          (X[i,]-centroids[j,])^2
                          )
                  })
                  )
    })
    return(idx)
}
 
computeCentroids <- function(X, idx, K) {
    centroids <- sapply(1:K, function(i) colMeans(X[idx == i,]))
    centroids <- t(centroids)
    return(centroids)
}
 
runkMeans <- function(X, K, max.iter = 10, plot=F, plot.progress=F) {
    initCentroids <- kMeansInitCentroids(X, K)
    K <- nrow(initCentroids)
    centroids <- initCentroids
    preCentroids <- centroids
    for (i in 1:max.iter) {
        idx <- findClosestCentroids(X, centroids)
        centroids <- computeCentroids(X, idx, K)
        if (plot.progress) {
            preCentroids <- rbind(preCentroids, centroids)
        }
    }
    xx <- data.frame(X, cluster=as.factor(idx))
    if(plot) {
        p <- ggplot(xx, aes(V1, V2))+geom_point(aes(color=cluster))
        if (plot.progress) {
            preCentroids <- data.frame(preCentroids,
                                       idx=rep(1:3, max.iter+1))
            p <- p+geom_point(data=preCentroids,
                              aes(x=V1, y=V2)) +
                geom_path(data=preCentroids,
                          aes(x=V1, y=V2, group=idx))
        }
        print(p)
    }
    return(xx)
}

After implemented this algorithm, I applied it to the dataset provided in ML-Class Ex 7.

?View Code RSPLUS
1
2
3
4
## dataset was converted from ML-class exercise 7  ex7data2.mat.
X <- read.delim("d:/ex7data2.txt", header=F)
K <- 3
xx <- runkMeans(X,K, plot=TRUE, plot.progress=TRUE)

The K-means code produced a visualization that steps the progress of the algorithm at each iteration.

Related Posts

  1. This is a wonderful work am very much impress in the way you normalized an iris data set to be well separated like this. can you please help me with the copy of your normalized iris data set, so as to use it in testing my own clustering algorithm too.
    Thanks.

    Reply

Leave a Comment


NOTE - You can use these HTML tags and attributes:
<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>