| BP网络模型的联合优化 |
| |
| 引用本文： | 崔丽群,刘万军,包明宇.BP网络模型的联合优化[J].辽宁工程技术大学学报(自然科学版),2004,23(Z1):81-82. |
| |
| 作者姓名： | 崔丽群 刘万军 包明宇 |
| |
| 作者单位： | 1. 辽宁工程技术大学,软件学院,辽宁,阜新,123000
2. 沈阳航空工业学院,机械与汽车学院,辽宁,沈阳,110034 |
| |
| 摘 要： | 针对 BP算法收敛速度慢、存在易于陷入局部极小值等缺点,不能有效地搜索到全局极小点。通过联合优化方法改进标准BP 算法,即通过黄金分割法动态调整学习系数:对 S 函数引入状态系数并进行分段;误差函数采用绝对和相对逼近精度相结合的方法。实验表明在一定程度上避免了学习中的局部极小问题,提高了学习效率,改进了网络的性能。
|
| 关 键 词： | BP 网络 改进算法 局部极小 联合优化 |
| 文章编号： | 1008-0562(2004)增刊-0081-02 |
| 修稿时间： | 2004年4月25日 |
Joint optimization of models of BP network |
| |
| CUI Li-qun,LIU Wan-jun,BAO Ming-yu.Joint optimization of models of BP network[J].Journal of Liaoning Technical University (Natural Science Edition),2004,23(Z1):81-82. |
| |
| Authors: | CUI Li-qun LIU Wan-jun BAO Ming-yu |
| |
| Institution: | CUI Li-qun~1 LIU Wan-jun~1 BAO Ming-yu~2 |
| |
| Abstract: | BP algorithm has the weaknesses such as slow convergent speed,easy getting into local minimum.So it is insurable to find global extreme value point.The joint optimization can improve standard BP algorithm as follows:dynamic adjusting learning rate by golden segment method;shape coefficient is introduced in sigmoid function and sectionalization;the copulative method of absolute and relative approach is adopted in error function. Experiment results show that it not only enhances the efficiency,but also can avoid getting into local minimum to some degree and modified the capacity of the network. |
| |
| Keywords: | BP network modified arithmetic local minimization Joint optimization |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
