The objective of this research is the presentation of a neural network capable of solving complete nonlinear algebraic systems of n equations with n unknowns. The proposed neural solver uses the classical back propaga...The objective of this research is the presentation of a neural network capable of solving complete nonlinear algebraic systems of n equations with n unknowns. The proposed neural solver uses the classical back propagation algorithm with the identity function as the output function, and supports the feature of the adaptive learning rate for the neurons of the second hidden layer. The paper presents the fundamental theory associated with this approach as well as a set of experimental results that evaluate the performance and accuracy of the proposed method against other methods found in the literature.展开更多
This project uses knowledge of neural network to analyze if the person under study is analyzed to be Parkinson disease patient or not. Binary classification is constructed based on the multi-feature database. A decisi...This project uses knowledge of neural network to analyze if the person under study is analyzed to be Parkinson disease patient or not. Binary classification is constructed based on the multi-feature database. A decision boundary is clearly plotted to separate patient with and without Parkinson disease. Results show that over 80% accuracy could be obtained with the preliminary results. Future efforts could be performed to construct more complicated neural network to improve the accuracy.展开更多
Spatial distribution of and interpolation methods for soil nutrients are the basis of soil nutrient management in precision agriculture.For study of application potential and characteristics of algebra hyper-curve neu...Spatial distribution of and interpolation methods for soil nutrients are the basis of soil nutrient management in precision agriculture.For study of application potential and characteristics of algebra hyper-curve neural network(AHCNN)in delineating spatial variability and interpolation of soil properties,956 soil samples were taken from a 50 hectare field with 20 m interval for alkaline hydrolytic nitrogen measurement.The test data set consisted of 100 random samples extracted from the 956 samples,and the training data set extracted from the remaining samples using 20,40,60,80,100 and 120 m grid intervals.Using the AHCNN model,three training plans were designed,including plan AHC1,using spatial coordinates as the only network input,plan AHC2,adding information of four neighboring points as network input,and plan AHC3,adding information of six neighboring points as network input.The interpolation precision of AHCNN method was compared with that of Kriging method.When the number of training samples was big,interpolation precisions of Kriging and AHCNN were similar.When the number of training samples was small,the precisions of both methods deteriorated.Since AHCNN method has no request on data distribution and it is non-linearization of neutron input variables,it is suitable for delineation of spatial distribution of nonlinear soil properties.In addition,AHCNN has an advantage of adaptive self-adjustment of model parameters,which makes it proper for soil nutrient spatial interpolation.After comparison of mean absolute error d,root mean squared error RMSE,and mean relative error%d,and the spatial distribution maps generated from different methods,it can be concluded that using spatial coordinates as the only network input cannot simulate the characteristics of soil nutrient spatial variability well,and the simulation results can be improved greatly after adding neighboring sample points’information and the distance effect as network input.When the number of samples was small,interpolation precision can be improved after properly increasing the number of neighboring sample points.It was also showed that evaluation of interpolation precision using conventional error statistic indexes was defective,and the spatial distribution map should be used as an important evaluation factor.展开更多
针对时差定位法受不同模式波速度差异及波形传播畸变等因素影响的问题,将神经网络技术应用到声发射源定位中。为了克服传统BP算法训练时间长和精度不够高的缺点,提出代数神经网络概念,在网络训练阶段引入代数算法,将复杂的非线性优化问...针对时差定位法受不同模式波速度差异及波形传播畸变等因素影响的问题,将神经网络技术应用到声发射源定位中。为了克服传统BP算法训练时间长和精度不够高的缺点,提出代数神经网络概念,在网络训练阶段引入代数算法,将复杂的非线性优化问题转化为简单的代数方程组求解问题,直接获得最优点,大大缩短了网络的学习时间。同时作为定位特征研究分析了转子碰摩声发射信号的非线性动力学特性,提出了关联维数、最大Lyapunov指数和K o lm ogorov熵等声发射源的非线性动力学新特征,并将其作为神经网络的输入定位特征。实验结果表明,利用这些声发射源的非线性动力学特征和神经网络结合能较好地解决了碰摩声发射源定位问题,为转子碰摩故障诊断提供依据,具有良好的应用前景和进一步研究的价值。展开更多
文摘The objective of this research is the presentation of a neural network capable of solving complete nonlinear algebraic systems of n equations with n unknowns. The proposed neural solver uses the classical back propagation algorithm with the identity function as the output function, and supports the feature of the adaptive learning rate for the neurons of the second hidden layer. The paper presents the fundamental theory associated with this approach as well as a set of experimental results that evaluate the performance and accuracy of the proposed method against other methods found in the literature.
文摘This project uses knowledge of neural network to analyze if the person under study is analyzed to be Parkinson disease patient or not. Binary classification is constructed based on the multi-feature database. A decision boundary is clearly plotted to separate patient with and without Parkinson disease. Results show that over 80% accuracy could be obtained with the preliminary results. Future efforts could be performed to construct more complicated neural network to improve the accuracy.
基金This research was financially supported by the National Natural Science Foundation of China(30600375)the National High Technology Research and Development Program of China(2006AA10A306,2006AA10Z271)。
文摘Spatial distribution of and interpolation methods for soil nutrients are the basis of soil nutrient management in precision agriculture.For study of application potential and characteristics of algebra hyper-curve neural network(AHCNN)in delineating spatial variability and interpolation of soil properties,956 soil samples were taken from a 50 hectare field with 20 m interval for alkaline hydrolytic nitrogen measurement.The test data set consisted of 100 random samples extracted from the 956 samples,and the training data set extracted from the remaining samples using 20,40,60,80,100 and 120 m grid intervals.Using the AHCNN model,three training plans were designed,including plan AHC1,using spatial coordinates as the only network input,plan AHC2,adding information of four neighboring points as network input,and plan AHC3,adding information of six neighboring points as network input.The interpolation precision of AHCNN method was compared with that of Kriging method.When the number of training samples was big,interpolation precisions of Kriging and AHCNN were similar.When the number of training samples was small,the precisions of both methods deteriorated.Since AHCNN method has no request on data distribution and it is non-linearization of neutron input variables,it is suitable for delineation of spatial distribution of nonlinear soil properties.In addition,AHCNN has an advantage of adaptive self-adjustment of model parameters,which makes it proper for soil nutrient spatial interpolation.After comparison of mean absolute error d,root mean squared error RMSE,and mean relative error%d,and the spatial distribution maps generated from different methods,it can be concluded that using spatial coordinates as the only network input cannot simulate the characteristics of soil nutrient spatial variability well,and the simulation results can be improved greatly after adding neighboring sample points’information and the distance effect as network input.When the number of samples was small,interpolation precision can be improved after properly increasing the number of neighboring sample points.It was also showed that evaluation of interpolation precision using conventional error statistic indexes was defective,and the spatial distribution map should be used as an important evaluation factor.
文摘针对时差定位法受不同模式波速度差异及波形传播畸变等因素影响的问题,将神经网络技术应用到声发射源定位中。为了克服传统BP算法训练时间长和精度不够高的缺点,提出代数神经网络概念,在网络训练阶段引入代数算法,将复杂的非线性优化问题转化为简单的代数方程组求解问题,直接获得最优点,大大缩短了网络的学习时间。同时作为定位特征研究分析了转子碰摩声发射信号的非线性动力学特性,提出了关联维数、最大Lyapunov指数和K o lm ogorov熵等声发射源的非线性动力学新特征,并将其作为神经网络的输入定位特征。实验结果表明,利用这些声发射源的非线性动力学特征和神经网络结合能较好地解决了碰摩声发射源定位问题,为转子碰摩故障诊断提供依据,具有良好的应用前景和进一步研究的价值。
