Abstract: The exponential synchronization problem of uncertain neural networks with the phenomenon of packet losses is studied. The proposed functional Lyapunov-Krasovskii can make full use of the system information. An improved integrating inequality is utilized to obtain the stability criterion formed in linear matrix inequalities and to solve the controller. The obtained result is of low conservatism, a larger attenuation rate can be realized. A switch system method is used to analyze the exponential stability of the system, the synchronization errorm system with packet losses is modeled as a standard switch system model. The mean residence time related to the packet loss rate is deduced. Finally, the effectiveness of the theory is verified by two numerical examples.