A regularization process to implement self-organizing neuronal networks

Kohonen Self-Organizing maps are interesting computational structures because of their original properties, including adaptive topology and competition, their biological plausibility and their successful applications to a variety of real-world applications. In this paper, this neuronal model is presented, together with its possible implementation with a variational approach. We then explain why, beyond the interest for understanding the visual cortex, this approach is also interesting for making easier and more efficient the choice of this neuronal technique for real-world applications