Just as the human brain learns to distinguish between different genres, melodies and instruments after listening to music for years, AI also has to listen for a time before it can analyse pieces of music with the help of certain parameters. There are two ways to do this: The “connectionist” (linking and multidimensional) models function similarly to the human brain. A network of nerve cells processes outside information by linking cells together. Why and how this functions is something we don’t yet know. The second method is based on a two-dimensional model known as the “Kohonen” map after its Finnish developer, Teuvo Kalevo Kohonen. The map illustrates how the neuron field organises itself during learning by placing similar stimuli close to each other – thus making learning transparent and allowing us to trace how the results are arrived at.
Kohonen, T. (2001). Self-Organizing Maps. Heidelberg: Springer.