Data Modeling: Self-Organizing Maps (2 of 2)
Objects can be characterized by numerous features. Self-Organizing Maps (SOM), or similarity maps, reduce these large number of features into a two dimensional map to ease interpretation. The SOM training algorithm is an unsupervised classification algorithm.
The example below uses nutritional information. The "SOM" image is simply the map of clusters of food, with food of similar nutritional content clustered near one another. Every subsequent image is one of the components of the nutrition. Red clusters are high values, indicating the presence of the feature. Dark blue indicates the absence. Intermediate colors show that cluster averages are somewhere in between.
See also the Bird Classification SOM.
Please be patient while the images load.