Stability analysis of delayed neural networks based on improved quadratic function condition

The stability of a class of neural networks (NNs) with time-varying delay is explored in this work. The characteristics of integral inequalities are considered, and the general stability condition (GSC) of delayed NNs avoiding high-order delay is given in the form of quadratic functions. For the GSC of NNs, under the number of decision variables remains unchanged, the information of time-delay and its derivatives are further excavated by delay-partitioning approach. Meanwhile, the free-moving points are established in each subintervals divided, and the traditional static constraints are transformed into dynamic constraints, which relaxes the feasibility space. Based on these methods, the improved stability criteria are given. Finally, two examples are provided to illustrate the meliority of the method under the same number of decision variables.