多层前向网络的研究--遗传BP算法和结构优化策略
Research on Multilayer Feedforward Neural Networks--Genetic Backpropagation Algorithm&Structure Optimization Strategy
-
摘要: 分析了引起BP算法局部收敛现象的原因,探讨了解决的途径.通过对遗传算法中 的基因群体的实数化,并将其与BP算法有机地集成起来,提出了遗传BP算法.此外,在借鉴 网络剪枝法思想的基础上,重新定义网络结构复杂度函数,并以遗传算法直接求解网络结构 优化问题,给出了网络结构优化策略.仿真结果和实际应用均表明了上述研究的成效.Abstract: In this paper, when classical genetic algorithms are applied to the region of real number and integrated with BP algorithm, a hybrid GA-BP learning algorithm is proposed. Besides, with the aid of genetic algorithms as well as the improvement of network complexity function, a structure optimization strategy which is on the basis of network stripping is proposed. The efficiency of research work mentioned above has been shown by numerical simulations and industrial applications.
计量
- 文章访问数: 3113
- HTML全文浏览量: 86
- PDF下载量: 1325
- 被引次数: 0