2-D and N-D SOMs

• A self organizing map (SOM) is a collection of nodes.
• Typically, these nodes are points in 2-D, or N-Dimensional space.
• Kohonen, who popularized the idea, says that it maps to the brain. There are indeed parts of the brain where brain topology corresponds with input (e.g. retinotopic and tonotopic maps, and indeed somatosensory cortex) but for our purposes that is off the point.
• You want quite a few nodes (dozens or even thousands).
• We'll do 2D space, but it's entirely straightforward to use larger dimensions.
•