Improved stability criteria for recurrent neural networks with interval time-varying delays via new Lyapunov functionals

This paper considers the stability problem of recurrent neural networks with interval time-varying delays. Based on a new augmented Lyapunov-Krasovskii functional that contains four triple integral terms and additional terms obtained from the activation function condition, a stability condition is derived in terms of linear matrix inequalities (LMIs). Also, a further improved stability criterion is derived by bounding the derivative of a special case of the proposed Lyapunov-Krasovskii functional based on a new inequality proposed in Seuret and Gouaisbaut (2013) [27]. A numerical example shows the improvement of the proposed approach both in terms of computational complexity and conservatism.