许少华;宋美玲;许辰;朱新宁;

• 1:东北石油大学计算机与信息技术学院

摘要(Abstract)：

针对过程神经网络(PNN)单一训练算法自适应调整能力差、缺乏对学习性质有效控制的问题,提出一种梯度下降与牛顿迭代相结合的求解算法——混合误差梯度下降算法.在训练初始阶段,基于网络训练目标函数,采用梯度下降法进行迭代寻优,只需计算目标函数一阶导数数值公式,复杂度低且误差下降快;当梯度下降法学习效率降低时,引入牛顿迭代法,并将梯度下降法的训练结果作为初始参数代入目标函数,使问题转化为求解非线性方程组,不需要一维搜索而提高网络训练效率.通过学习效率分析自适应调节两种算法的切换,直至满足停机条件.将其应用于时变信号模式分类,实验结果表明,该算法较大地提高PNN训练效率.

关键词(KeyWords)：过程神经元网络;算法效率;牛顿迭代法;梯度下降法;混合误差梯度下降算法

Abstract:

Keywords:

基金项目(Foundation):国家自然科学基金项目(61170132)

作者(Author):许少华;宋美玲;许辰;朱新宁;

Email:

参考文献(References)：

• [1]何新贵,梁久祯.过程神经元网络的若干理论问题[J].中国工程科学,2000,2(12):40-44.He Xingui,Liang Jiuzhen.Some theoretical issues on process neural networks[J].Engineering Science,2000,2(12):40-44.

• [2]许少华,何新贵.关于连续过程神经元网络的一些理论问题[J].电子学报,2006,34(10):1838-1842.Xu Shaohua,He Xingui.Some theoretical issues on continuous process neural networks[J].Acta Electronica Sinica,2006,34(10):1838-1841.

• [3]刘显德,刘立伟,许少华.一种过程神经元网络模型及其在动态预测中的应用[J].大庆石油学院学报,2008,32(4):107-110.Liu Xiande,Liu Liwei,Xu Shaohua.A process neural networks model and its application to dynamic forecasting[J].Journal of Daqing Petroleum Institute,2008,32(4):107-110.

• [4]许少华,王皓,王颖.基于过程神经网络的多维动态信息处理技术[J].大庆石油学院学报,2010,34(5):145-164.Xu Shaohua,Wang Hao,Wang Ying.Multidimensional dynamic information processing based on process neural networks[J].Journal of Daqing Petroleum Institute,2010,34(5):145-164.

• [5]许少华,何新贵.基于函数正交基展开的过程神经元网络学习算法[J].计算机学报,2004,7(5):645~650.Xu Shaohua,He Xingui.Learning algorithms of process neural networks based on orthogonal function basis expansion[J].Chinese J of Computers,2004,27(5):645-649.

• [6]许少华,肖红.基于离散Walsh变换的过程神经元网络学习算法[J].大庆石油学院学报,2003,27(4):58-61.Xu Shaohua,Xiao Hong.Learning algorithm of process neural networks based on discrete Walsh conversion[J].Journal of Daqing Petroleum Institute,2003,27(4):58-61.

• [7]Kaveh A,Talatahari S.Particle swarm optimizer,ant colony strategy and harmony search scheme hybridized for optimization of truss structures[J].Computers&Structures,2009,87(5):267-283.

• [8]Ababu Teklemariam Tiruneh.Higher order aitken extrapolation with application to converging and diverging Gauss-Seidel iterations[J].Journal of Applied Mathematics and Physics,2013,1(5):128-143.

• [9]Zhong S S,Li Y,Ding G.Continuous wavelet process neural network and its application[J].Neural Network World,2007,17(5):483-495.

• [10]Li Ge,Yin Guisheng.Application of double hidden layer process neural network on wood drying modeling[C]∥Proceedings of 2011IEEE the 18th International Conference on Industrial Engineering and Engineering Management(Volume 3).Bejing:Electronic Industry Press,2011.

• [11]丁刚.过程神经网络泛化能力的理论与应用研究[D].哈尔滨:哈尔滨工业大学,2008.Ding Gang.Generalization capability of process neural network and its applications[D].Harbin:Harbin Institute of Technology,2008.

• [12]刘坤,谭营,何新贵.基于粒子群优化的过程神经网络学习算法[J].北京大学学报,2011,47(2):238-244.Liu Kun,Tan Ying,He Xingui.Particle swarm optimization based learning algorithm for process neural networks[J].Acta Scientiarum Naturalium Universitatis Pekinensis,2011,47(2):238-244.

• [13]荆献勇,肖明清,余文波.改进免疫算法在预测过程神经网络中的应用[J].系统工程与电子技术,2010,32(10):2136-2140.Jing Xianyong,Xiao Mingqing,Yu Wenbo.Application of improved immune algorithm in forecast procedure neural network[J].Systems Engineering and Electronics,2010,32(10):2136-2140.

• [14]葛利.一种基于混合遗传算法学习的过程神经网络[J].哈尔滨工业大学学报,2005,37(7):986-988.Ge Li.Procedure neural networks based on being learned by hybrid genetic algorithm[J].Journal of Harbin Institute of Technology,2005,37(7):986-988.

• [15]李欣,程春田.基于过程神经网络和量子免疫算法的油气评价[J].计算机工程与应用,2008,45(25):9-13.Li Xin,Cheng Chuntian.Evaluation method of oil and gas resource based on process neural network and quantum immune algorithm[J].Computer Engineering and Applications,2008,45(25):9-13.

• [16]关治,陆金普.数值分析基础[M].北京:高等教育出版社,1998.Guan Zhi,Lu Jinpu.Numerical analysis[M].Beijing:China Higher Education Press,1998.

• [17]王能超.同步并行算法设计[M].北京:科学出版社,1996.Wang Nengchao.Design of synchronous parallel algorithm[M].Beijing:Science Press,1996.