摘要
利用一种改进粒子群优化算法的粒子速度、位置修正公式替代传统BP算法的梯度下降法;并基于改进BP神经网络建立了电磁兼容问题中导线串扰电压的预测模型;实验结果表明改进后的方法在预测精度和小样本性能方面要比传统BP神经网络高得多.
In this paper,particle velocity and position correction formula in improved particle swarm optimization was usedinstead of gradient descent in traditional BP algorithm;and prediction model of crosstalk voltage in EMC was established based onimproved BP neural network;Experimental results showed that forecasting accuracy and small sample performance of improvedmethod was much higher than the traditional BP neural network.
出处
《晋中学院学报》
2010年第3期8-11,23,共5页
Journal of Jinzhong University
基金
晋中学院教改资助项目(JG20090104
JG20090119
JG20090205)
山西省教育厅资助项目(山西省普通本科高等教育教学改革项目
晋教高[2009]19号)
