Self-Organizing Maps (SOM) are a class of unsupervised learning neural networks that realize a nonlinear projection of high-dimensional data onto a low-dimensional, usually two-dimensional grid structure.
The model was developed by Teuvo Kohonen and is based on a competitive learning process, in which for each input the best matching neuron (Best Matching Unit, BMU) is determined.
The weights of the BMU and its topological neighbors are gradually adjusted to map the input data while preserving structure.
This method allows the visualization of data clusters and the detection of latent structures without explicit class labels.
Definition:
“A self-organizing map is an unsupervised neural network that produces a topological mapping of high-dimensional input data onto a low-dimensional grid structure through competitive learning.”
Source:
Kohonen, T. (1982). Self-Organized Formation of Topologically Correct Feature Maps. Biological Cybernetics, 43(1), 59-69.