两种SA-BP融合方式性能比较分析
comparative analysis of two fusion manners of SA-BP
模拟退火（SA）和误差反传算法（BP）常常融合成混合算法用于神经网络的学习。本文针对SA与BP算法的两种融合方式：即先BP后SA学习（SA-BP I ）与先SA后BP学习（SA-BP II），进行了性能比较分析。在参数设置相同的情况下，通过神经网络逼近非线性函数，对两种融合方式的学习性能进行了仿真研究。结果表明，SA-BP II具有更好的精度和鲁棒性。
Simulated annealing (SA) and error back propagation (BP) algorithm is often fusing into a hybrid learning algorithm for neural networks training. In this paper, the performance of two kinds of algorithm fusion manners, first BP then SA（SA-BP I ）and SA first then BP（SA-BP II）, is compared. The simulated study of two manners' performance is implemented by approximating a nonlinear function with same parameters setting.The results show that SA-BP II has better accuracy and robustness.
於世为、诸克军
自动化基础理论计算技术、计算机技术
模拟退火BP 算法融合性能比较
Simulated annealingback propagation algorithmfusionperformance comparison
於世为,诸克军.两种SA-BP融合方式性能比较分析[EB/OL].(2011-06-01)[2025-08-16].http://www.paper.edu.cn/releasepaper/content/201106-16.点此复制
评论