传统的联想记忆神经网络模型是根据联想记忆点设计权值．文中提出一种根据联想记忆点设计基于约束区域的神经网络模型，它保证了渐近稳定的平衡点集与样本点集相同，不渐近稳定的平衡点恰为实际的拒识状态，并且吸引域分布合理．它具有学习和遗忘能力，还具有记忆容量大和电路可实现优点，是理想的联想记忆器． The traditional associative memory neural network model is based on associative memory point design weights. In this paper, a neural network model based on associative memory points based on constrained regions is proposed, which guarantees that the asymptotically stable equilibrium point set is the same as the sample point set, and the asymptotic stable equilibrium point is just the actual rejection state and attracts Domain distribution is reasonable. It has the learning and forgetting ability, but also has the memory capacity and circuit can realize the advantages, is the ideal associative memory.