From entry clustering in glossary Artificial Intelligence: humans at the heart of algorithms
Clustering algorithms take data and tries to sort it into clusters each with some form of similar characteristics. This is like you might do if looking a scatter graph and noticing that the data falls into two or three groups. Examples of cluster analysis includes k-means which is based around calculating centroids of each cluster and Kohonen networks (or self-organsing maps) which are a form of neural netwrok where similar data points are classified close to one another on a two dimensional grid. Clustering algorithms are a form of unsupervised learning, but there are also semi-supervised forms where some or all of the training set has a classification and the clustering algorithm tries to ensure that each cluster has a consistent label, even if there may be several clusters for each label.