Abstract: In order to simulate the chaos phenomenon of the neural network, the phase space reconstruction technique is expounded, and the method of determining the largest Lyapunov exponent and the correlation dimension of the system from a time series is introduced. By utilizing the Lyapunov exponent as criterion, the strange attractors of the three-layer feedback neural network are constructed, the activity characteristics of the strange attractor are analyzed, and the correlation dimensions of the strange attractors are calculated. The results show that the chaos neural network has the complex dynamic characteristics, while it has all kinds of attractors, not only the fixed point,the limit cycle, and the tori, but also the strange attractors.