摘要
传统前馈神经网络盲均衡中神经网络的初始权重的确定缺乏理论依据,收敛速度慢,容易陷入局部极小值.为有效克服这些缺陷,提出了遗传优化神经网络的盲均衡算法.算法用遗传算法对前馈神经网络的网络权重进行优化,为神经网络提供一个全局较优的局部搜索空间;再利用传统神经网络在这个局部空间进行更精确地搜索,最终实现盲均衡.计算机仿真结果表明:与传统神经网络算法相比,新算法达到了更好的收敛特性和均衡效果,剩余稳态误差减少30%以上,收敛速度加快约20%,误码率也有明显降低.
There are some disadvantages such as slow convergence, easy local minimum and no enough theoretical evidence for determining the initial weight of neural network by the traditional FNN blind equalization algorithm. Therefore a genetic algorithm optimizing neural network （GA-BP） is proposed to overcome these disadvantages. First, a preferable local solution space is offered to the neural network by using GA. Then, a precise searching is realized in the space with the conventional neural network algorithm to finish the blind equalization. Finally the comparisons between the traditional FNN algorithm and the GA-BP are carried out, and the simulating results show that the remainder steadystate error has decreased by more than 30%, the convergence rate has improved by 20% and the bit error rate（BER） has dropped obviously.
出处
《中北大学学报（自然科学版）》
CAS
北大核心
2009年第2期137-142,共6页
Journal of North University of China(Natural Science Edition)
关键词
盲均衡
神经网络
BP算法
遗传算法
blind equalization
neural network
BP algorithm
genetic algorithm
作者简介
李沅（1982-），女，助教，硕士．主要从事信号与信息处理方面的研究．
