In class, we covered the simple Bayesian classifier. This form of classification falls under a category known as supervised learning. What this means is that a set of labelled data data is used to to "train" the underlying model. However, it is not always possible to have such a data set, yet we may still wish to discover some form of underlying structure in an unlabelled data set. Such a task falls under the category of unsupervised learning.

Clustering is a form of unsupervised learning. "Clustering is the problem of identifying groups, or clusters of data points in multidimensional space". For example, consider the following data set:

Figure 1

Although this data set is not labelled, we can still clearly see three different groups, or clusters, of data points, as shown here:

Figure 2

Back to ECE662, Spring 2014