摘要
论文在阐明了遗传算法和神经网络结合的必要性之后,分析了一般遗传算法在神经网络结构优化过程中存在的不足,并根据多物种之间相互竞争和相互适应的机理提出了一种基于多物种协同进化的优化方法。该方法既可以有效地避免神经网络结构寻优过程中解搜索空间过大以及进化规则复杂等问题,还可以起到对网络的结构和权值同时进化的作用。仿真实验表明该方法是可行并且有效的。
After demonstrating the necessity of combining Genetic Algorithm(GA) and Neural Networks (NN),this paper points out the shortcomings of the common NN structure optimization method based on GA,and gives a new method based on the species coevolution.The new method can not only avoid the matter of the large range of solution and the complexity of evolution rules effectively during NN structure learning,but also realize that the structure and weights of NN can be optimized together.The simulation and application demonstrate that this method is effective.
出处
《计算机工程与应用》
CSCD
北大核心
2006年第15期69-71,共3页
Computer Engineering and Applications
关键词
神经网络
遗传算法
多物种协同进化
机械手
Neural Networks,Genetic Algorithms, species coevolution,manipulator