摘要
针对广义回归网络的模式层单元数目与样本数量成正比的问题,提出了基于相似度衡量的模糊均值聚类的样本精简方法.针对广义回归网络在时变环境下难以确定平滑因子,自适应能力弱的缺点,提出了一种基于贡献率的选择优化方案.仿真结果表明,改进后的GRNN有较快的处理速度和较强的自适应能力,能够在实际应用中很好地辨识较为复杂的非线性时变系统.
For the defect that pattern layer neurons are proportional to the number of the training samples, the method of fuzzy means clustering based on similarity index to decrease samples is proposed. For the defect that it is hard to determine the smoothing parameter in time-varying conditions, the adaptive optimizing strategy based on contributing ratio is proposed. The simulation results show that the improved GRNN has fast solving speed and good adaptability. It can approach complex nonlinear time-varying systems well.
出处
《微电子学与计算机》
CSCD
北大核心
2009年第6期32-35,共4页
Microelectronics & Computer
