对多层前向小世界神经网络的网络参数、权值修正策略以及网络结构进行改进,提出一种基于层连优化的小世界神经网络的改进算法.通过对比现有各种不同形式的小世界神经网络,验证了上述改进的必要性.仿真结果表明,改进模型比现有小世界神经网络收敛速度更快,逼近精度更高,模型稳定性更强. In this paper, an improved algorithm of small-world neural network based on layer-by-layer optimization is proposed to improve the network parameters, weight correction strategies and network structure of multi-layer forward small-world neural networks. By comparing existing small-world Neural network to verify the necessity of the above improvement.The simulation results show that the improved model is faster than existing small-world neural networks in convergence rate, the approximation accuracy is higher and the model is more stable.