Using the Generation5 library I have ported part of this applet (Colour dataset) into a processing.

Kohonen network is a good example of unsupervised networks. It is quite simple yet introduces the concepts of self-organization and unsupervised training easily.

Kohonen network is often called self-organizing map (SOM) or self-organizing feature map (SOFM).

There are quite a few types of self-organizing networks, like the Instar-Outstar network, the ART-series, and the Kohonen network. For purposes of simplicity, we will look at the Kohonen network.

The Kohonen network is an n-dimensional network, where n is the number of inputs. For simplicity, we will look at a 2-dimensional networks. 1D, 2D and 3D networks are used the most.

Pic1.: Organization of the 2D Kohonen network [taken from]

Demo I have created contains 2D Kohonen network with dimensionality 40x40x3. It means. Size X = 40 neuron, Size Y = 40 neurons, input vector X consists from X1, X2, X3. The network is feeded with RGB colours from a learning dataset. Input vector is normalized from <0, 255> into <0.0, 1.0>. Weights are randomized before the learning, so the result after learning is always a little bit different - the layout of the learned clusters is different.

Demo

Wait while the network is trained. Once it is trained, set the RGB value and press the "Find" button. It will see in which cluster was your color assigned.