In this chapter, we've seen how to apply unsupervised learning algorithms on neural networks. We've been presented a new and suitable architecture for that end, the self-organizing maps of Kohonen. Unsupervised learning has proved to be as powerful as the supervised learning methods, because they concentrate only on the input data, without need to make input-output mappings. We've seen graphically how the training algorithms are able to drive the weights nearer to the input data, thereby playing a role in clustering and dimensionality reduction. In addition to these examples, Kohonen SOMs are also able to classify clusters of data, as each neuron will provide better responses for a particular set of inputs.