Evolutionary mechanisms in self-organizing map

In this paper a new model of self-organizing neural networks is proposed, which has realized the ability to project data from high dimensional space into low dimensional space, as well as to facilitate the visual inspection in the inherent topological structure of the projected data. In this model, a series of evolutionary working mechanisms of neurons have been introduced to overcome the weakness of conventional learning algorithm. The empirical evidences show the proposed algorithm is adaptive and robust. This work can be viewed as a development of classic Kohonen self-organizing feature map.