节点文献
几种改进BP算法在气液两相流流型识别中的比较
Comparison of Several Fast Learning Algorithms of BP Networks in Flow Pattern Identification of Gas-Liquied Two-Phase Flow
【Author】 Sun Bin Zhou Yunlong Hong Wenpeng (Department of Thermal Power, Northeast China Institute of Electric Power Engineering, Jilin 132012, China)
【机构】 东北电力学院动力系;
【摘要】 通过流型识别的实例对几种具有代表性的用以训练BP网络的改进算法进行性能对比研究。首先分析了基于标准梯度下降法和基于标准数值优化方法获得的各种改进算法的优缺点,然后对各种改进算法在训练中所需要的收敛时间及其达到的误差进行对比分析,为在BP网络识别流型时选择算法提供一些借鉴。
【Abstract】 A comparative study on several typical fast learning algorithms of BP networks is proposed by the examples of flow pattern recognition. Firstly the advantages and disadvantages of faster algorithms that based on the method of standard steepest and the method of standard numerical optimization are analyzed. Secondly the convergent time and the error of the faster algorithms are compared. Some references for selecting flow pattern identification algorithms of BP network are provided.
【Key words】 Standard steepest Numerical optimization Learning velocity Flow pattern identification;
- 【会议录名称】 第二届全国信息获取与处理学术会议论文集
- 【会议名称】第二届全国信息获取与处理学术会议
- 【会议时间】2004-08
- 【会议地点】中国大连
- 【分类号】TP183
- 【主办单位】中国仪器仪表学会