Scribd
Upload
76 views

LAB 6A:K-Means Clustering

K-means clustering aims to partition n observations into k clusters by minimizing the within-cluster sum of squares. It works by randomly assigning observations to initial clusters and calculating cluster means, then iteratively reassigning observations to clusters based on distance to cluster means until convergence. The document provides an example applying k-means clustering to classify 8 two-dimensional vectors into 2 clusters, initializing cluster means as c1 = (1 1)T and c2 = (4 0)T. A MATLAB application is described that allows users to perform k-means clustering on user-defined or predefined point sets in 2D space and visualize the results.

Uploaded by

Ann
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
76 views

LAB 6A:K-Means Clustering

K-means clustering aims to partition n observations into k clusters by minimizing the within-cluster sum of squares. It works by randomly assigning observations to initial clusters and calculating cluster means, then iteratively reassigning observations to clusters based on distance to cluster means until convergence. The document provides an example applying k-means clustering to classify 8 two-dimensional vectors into 2 clusters, initializing cluster means as c1 = (1 1)T and c2 = (4 0)T. A MATLAB application is described that allows users to perform k-means clustering on user-defined or predefined point sets in 2D space and visualize the results.

Uploaded by

Ann
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

LAB 6A:K-Means Clustering

1. Theoretical basis

1.1. Objective: A set of N points, described by vectors of size p must be


divided into K classes (groups, clusters)

The division of the set X into K classes is done so as to minimize the objective
function J, defined by:
K Nk
J ck ||2
k= 1 i =1
where :
- Nk = number of vectors in class k
- K = number of classes: k = 1 .. K
- ck, is the mean (center) of the class k
- Xi(k) = vector Xi belonging to class k

1.2. Algorithm:
1. The N vectors are randomly assigned in K classes and the class means (centers) are
calculated.
Another option is the random initialization of the centers (averages) of the K classes.
2. For each of the N vectors the distance from the centers of the classes is calculated
and the vector Xi is assigned to the class k for which the distance ||Xi – ck|| is minimum.
3. After traversing the entire set of vectors (p-dimensional spatial points) and re-
labeling all vectors, recalculate the centers (means) of the classes and resume the
algorithm from step 2.
The algorithm stops when no re-labeling of any vector occurs.
1.3. Problem
Consider the vectors :
A = (2 0)T, B = (3 1)T, C = (4 0)T, D = (3 -1)T
E = (-2 0)T,F = (-3 1)T, G=(-4 0)T, H=(-3 -1)T
It is desired to classify these vectors into K = 2 classes using K-Means Clustering
assuming the class averages are initialized with c1 = (1 1)T si c2 = (4 0)T.

2. Lab Application
Objective : Using the Unsupervised K-Means algorithm to classify two-dimensional
vectors into K-classes.
Vectors can be chosen as graphical points by the user or there can be used predefined
sets of points.
- Open the Matlab program and run the L6_Kmeans.m file.
- to add a new point write the coordinates of the point and then click Adauga
- to enter the number of classes, fill in the field Nr. Clase with the desired number of
classes.
By default there are 3 classes.
- To classify, click Clustering K-means
- If you want to use a predefined set of points, choose one of the predefined set of points
and then click Clustering K-means
- To delete all entered points, press Reset
- To delete the classification without deleting the points, click Clear plot
The number of iterations performed as well as the objective function J are displayed
in the Command Window.
The application looks like the one on the screen below
- To delete the classification without deleting the points, click Clear plot
The number of iterations performed as well as the objective function J are displayed in
the Command Window.
The application looks like the one on the screen below.

Fig. 1. MATLAB application of K-means Clustering to classify the 2-D vectors in K classes.

You might also like