Kohonen SOM: Unsupervised learning that preserves topological structure of high-dimensional data
Training Statistics
Epoch:0
Current LR:0.5
Current Radius:10
Quantization Error:-
Self-Organizing Maps (Kohonen networks) map high-dimensional data to a 2D grid while preserving topological relationships. Nearby neurons respond to similar inputs. The BMU (Best Matching Unit) and its neighbors are updated toward each training sample.