Chaos control of a class of learning algorithms for ICA neural networks

Independent component analysis (ICA) neural network is an important method for separating the mixed signal into independent components. However, the ICA neural networks can produce complicated dynamical behavior under certain conditions, such as the periodic oscillation, bifurcation and chaos. This paper introduces the chaos control of a class of ICA learning algorithm, Hyvarinen and Oja´s ICA namely, and the stability transformation method (STM) of chaos feedback control is utilized to the convergence control of the ICA. Time series diagrams, Lyapunov exponent of dynamical system demonstrate that the desired fixed points of iterative map of the ICA can be captured, and the chaotic behavior of the algorithm can be controlled.