论文部分内容阅读
针对BP训练方式采用梯度法易导致局部收敛的不足,提出一种融合进食粒子群算法(EPSO)和梯度法的Elman网络优化方法.首先,通过模拟鸟群进食行为得到一种EPSO算法,以改善标准PSO的全局性能;然后,将EPSO用于Elman网络权值的全局优化,同时将梯度法用于EPSO的进食过程局部搜索,以提高解的局部收敛性能;最后,将该网络优化方法用于飞行轨迹预测实验,仿真结果表明了其有效性.
Aiming at the insufficiency of local convergence caused by gradient method in BP training method, an Elman network optimization method based on feed particle swarm optimization (EPSO) and gradient method is proposed.Firstly, an EPSO algorithm is obtained by simulating bird feeding behavior to improve Then the global optimization of the weights of Elman network is applied to EPSO, and the gradient method is applied to the local search of food process of EPSO to improve the local convergence of the solution. Finally, the method of network optimization is applied to Flight trajectory prediction experiment, simulation results show its effectiveness.