Cluster and Distance Measure

Synonyms

Segmentation; Unsupervised learning

Definition

Clustering

Clustering is the assignment of objects to groups of similar objects (clusters). The objects are typically described as vectors of features (also called attributes). So if one has n attributes, object x is described as a vector (x₁,..,x_n). Attributes can be numerical (scalar) or categorical. The assignment can be hard, where each object belongs to one cluster, or fuzzy, where an object can belong to several clusters with a probability. The clusters can be overlapping, though typically they are disjoint. Fundamental in the clustering process is the use of a distance measure.

Distance Measure

In the clustering setting, a distance (or equivalently a similarity) measure is a function that quantifies the similarity between two objects.

Key Points

The choice of a distance measure depends on the nature of the data, and the expected outcome of the clustering process. The most important consideration is the type of the features...