A self-organizing map (SOM) is a competitive learning network (an interesting class of unsupervised machine learning), and it is one of the most popular neural network models. In this network, only one neuron gets activated at a given time, so the output neurons compete among themselves to be activated. This activated neuron is called the winning neuron. When one neuron fires, its closest neighbors tend to get more excited than ones that are further away (defining a topological neighborhood with decaying distance). As a result, the neurons are forced to organize themselves (through an adaptive or learning process) and a feature map between inputs and outputs is created. That's why this network is known as a self-organizing map.
The adaptive process of the SOM algorithm takes place in the following two steps:
- Ordering (self-organizing...
