白噪声混沌神经网络的模拟退火策略

Simulated annealing strategy in white noise chaotic neural network model

分析了白噪声混沌神经网络模型的动力学特性和对自反馈连接权值的敏感性,研究了退火函数在优化过程中对准确性和计算速度的影响.利用分段模拟退火思想对白噪声混沌神经网络进行改进,使得该网络模型在保证优化算法准确性的基础上,加快了收敛速度,并通过对经典旅行商问题的仿真实验,表明算法具有很强的克服陷入局部极小点的能力,较大程度地改善了原模型的求解组合优化问题的能力,验证了这种分段模拟退火策略的有效性.最后说明了模型参数对改进网络性能的重要性.

This paper analyzes the dynamics characteristic of white noise chaotic neural network model and the sensitivity to self-feedback connection weight,studies the annealing function in the optimized process to accurate and the computation speed influence.Using the sub-annealing thought improves the white noise chaotic neural network,and makes this network model under the basis on the guarantee optimization algorithm accurate,speed up the convergence rate,and using the classical traveling salesman（TSP） study indicate the algorithm has the very strong capacity to overcome the local minimum point.The greater improves the ability of solving combination optimization question,confirms this kind of sub-annealing strategy validity.The results show the importance of the model parameters to improve the performance of the network.