How do self-organizing maps (SOMs) work in modern unsupervised systems?

SOMs project high-dimensional data onto a low-dimensional grid while preserving topology.

Self-Organizing Maps are neural networks that map high-dimensional inputs onto a 2D grid. Neurons compete to represent input data, and neighbors in the grid are updated together, preserving topological relationships. Although less common in deep learning today, SOMs are still used for visualization and clustering interpretability in unsupervised systems.