A self-organising map (SOM) is a form of unsupervised learning where nodes on a 2D grid (or more geometric lattice structure) act as archetypes or key values for a form of clustering, where like items are allocated to grid points that have some form of relationship. The prime example of a SOM are Kohonen networks. Like other forms of clustering algorithm, SOM can be used for data interpretation where a human expert labels areas on the map, or as a form of dimension reduction either using the nodes as discrete categorical clusters or using the 2D (or higher dimensional) embedding of the grid or lattice as continuous dimensions.
Used in Chap. 5: page 60; Chap. 6: pages 74, 81, 82, 84; Chap. 16: page 241; Chap. 21: page 339
Also known as self-organizing map, self-organising network, self-organizing network