A Fractal Selforganizing Map with Partially Chaotic Neurons

A new type of a chaotic neuron is proposed which allows a new, simplified architecture of selforganizing maps. The chaotic neurons are examined with respect to their application in selforganizing maps. A new concept of self-organizing maps with fractal architecture is proposed and its suitability for a VLSI-implementation is examined. It is shown that the efficiency of conventional selforganizing maps can be overcome while the expenditure, with regard to a hardware-realisation, is notably reduced.