Improved Stability Criterion for Delayed Neural Networks via Negative-Definiteness Condition of Quadratic Function

This paper concerns a stability analysis problem of delayed neural networks via the negative-definiteness condition of quadratic function. Firstly, we construct the new augmented Lyapunov-Krasovskii functionals (LKFs) which provide the quadratic form of delays to the derivative of the LKFs to reflect additional delay information. Secondly, the integral inequality that considers the cross information between augmented state terms is utilized as a bounding technique for estimating the upper bound of the integral terms in the derivative of the LKFs. Finally, an improved stability criterion is obtained by using the quadratic function negative-definiteness technique on the upper bound of the LKF derivative. The conservatism of the derived stability criterion is evaluated by comparing with existing results by well-known numerical examples.