本文提出并研究了一种具有全局收敛特性的神经网络模型。该模型的动态特性可由对偶微分方程描述,它具有从状态空间内任一初始点找出多个稳定平衡点的能力。在应用方面,它很容易发展为诸如解决全局优化问题等要求全局收敛特性的专用神经网络。 This paper presents and studies a neural network model with global convergence. The dynamic properties of the model can be described by a dual differential equation with the ability to find multiple stable equilibrium points from any initial point in the state space. On the application side, it is easy to develop a special neural network that requires global convergence, such as solving global optimization problems.