SOM Training

• Based on rewarding the winning node

• This is a form of competitive learning

• Winners weights are adjusted to be closer to the input vector

• Why not equal?

• We want the output map to learn regions, not examples