一种基于参数扰动退火策略的神经网络全局优化新算法

A New Algorithm Based on Hopfield Neural Network with the Parameter Disturbances Simulated Annealing for the Global Optimization

针对非线性全局优化问题 ,提出了一种新算法 :它采用参数扰动策略 ,使 Hopfield神经网络克服局部极值点的吸引 ,同时对参数扰动采用模拟退火算法 ,使扰动逐渐减小 ,直到扰动不能对最优解和最优值产生影响 ,从而得到全局最优解。通过对大量测试函数的仿真计算 ,充分体现了新算法在速度、精度和适应性方面的优势。本文还对算法的收剑性进行了理论分析。

To solve the global optimization of nonlinear problems, this paper proposes a new algorithm, which makes the Hopfield neural network escape from the attraction of the local minimum points by the strategy of parameter disturbances. By using the simulated annealing algorithm, the parameter disturbances will be gradually reduced until it can not affect the network converging to the global optimization solutions. Therefore, we can finally get the global optimization of the objective function. The advantages of the new algorithm are that the present algorithm possesses rapidness, accuracy, and robustness, which are demonstrated by the experimental results of several standard test problems. At the same time, the convergence of the new algorithm is also analyzed.