LMI based global asymptotic stability criterion for recurrent neural networks with infinite distributed delays

Abstract

Global asymptotic stability problem for a class of recurrent neural networks with infinite distributed delay is investigated based on the linear matrix inequality (LMI) technique. Using a matrix decomposition method, a vector-matrix form of recurrent neural networks with infinite distributed delay is obtained. Then by constructing a suitable Lyapunov functional and using an inequality, new LMI-based criteria are established to ensure the global asymptotic stability of the class of neural networks, which considers the effects of neuron's excitatory and inhibitory action in the term of infinite delay on the networks. The obtained results are independent of the size of delay and are easily verified. Numerical example shows the effectiveness of the obtained results. © 2009 Springer Berlin Heidelberg.