Labeling a set of observations for classification or regression can be a daunting task, especially in the case of a large feature set. In some cases, labeled observations are either not available or not possible to create. In an attempt to extract some hidden association or structures from observations, the data scientist relies on unsupervised learning techniques to detect patterns or similarity in data.

The goal of unsupervised learning is to discover patterns of regularities and irregularities in a set of observations. These techniques are also applied in reducing the solution space or feature set similarly to the divide-and-conquer approach commonly used in Computer Science.

There are numerous unsupervised algorithms; some are more appropriate to handle dependent features while others generate more relevant groups in the case of hidden features [4:1]. In this chapter, you will learn three of the most common unsupervised learning algorithms: