基于动态调整位移参数的BP改进算法  

BP Improvement Algorithm Based on Random Parameters Adjustment

作　　者：李森林[1] 邓小武[1] 

机构地区：[1]怀化学院计算机科学与技术系,湖南怀化418008

出　　处：《邵阳学院学报（自然科学版）》2013年第2期18-21,共4页Journal of Shaoyang University：Natural Science Edition

摘　　要：人工神经网络应用中,80%～90%采用BP网络,BP神经网络实质是一个无约束非线性最优化计算过程,计算时间长,且难得到最优结果.文中提出了一种动态调整位移参数的BP改进算法,使得BP网络能尽快跳出平坦区,加快计算速度.实验对太阳黑子进行预测,证明改进后BP算法具有速度快、精度高等方面的优点,达到了预期效果.BP network is adopt in neural networks application about 80% - 90% ,it is a unrestraint nonlinearity optimization calculation process. For calculation time being long and optimum result being difficulty, a modified BP algorithm based on random parameters adjustment is proposed, which is easy to jump out of even area and accelerate a computing speed. Final, we test the improved algorithm with data of the macula in VC platform, and analyze these data. We find that the improved algorithm has fast convergence and high accuracy.

关 键 词：反向学习算法 参数调整 神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]