Solving Unsupervised Learning Problems

Unsupervised learning problems, when no output variable is present, seek to find some kind of structure in the underlying data, such as groups or clusters of attributes, according to their similarity: two examples belonging to the same group must exhibit a higher value of similarity than two patterns associated with different clusters.

Techniques related to unsupervised learning are usually called clustering algorithms. The number k of clusters may be chosen initially by the user (e.g. in the k-means technique) or suggested by the algorithm.

Available unsupervised learning tasks