A new algorithm for chaotic system identification based on self-organizing neural gas

This paper presents a novel algorithm, called the self-growing neural gas network (SGNGN) method, for chaotic system identification. Combined with local linearization of the system state space, the proposed method can be applied to identify chaotic system or predict chaotic time series. Compared to the neural gas network, the SGNGN method allows the population of neurons to grow with the presentation of input vectors. Simulations results show that the proposed method can greatly accelerate the processing of convergence of weights