小波网络具有小波的多尺度特性和神经网络的自学习功能,在回归估计中得到广泛的应用,但其性能受到样本中粗差的严重影响.虽然以M-估计作为目标函数可以解决这个问题,但由于其对应的影响函数由残差绝对值决定,因此如何选择初始参数值成...小波网络具有小波的多尺度特性和神经网络的自学习功能,在回归估计中得到广泛的应用,但其性能受到样本中粗差的严重影响.虽然以M-估计作为目标函数可以解决这个问题,但由于其对应的影响函数由残差绝对值决定,因此如何选择初始参数值成为一个关键问题.为此,提出回归函数的小波支持向量机鲁棒估计方法(小波支持向量回归,WSVR,Wavelet Support Vector Regression).该方法中首先提出并证明了一种新的小波支持向量机(WSVM,Wavelet Support Vector Machine),用于确定初始参数值方法,这种方法能够确定合理的网络结构和合适的初始参数值,保证含有粗差的样本点的残差绝对值较大;然后使用一种构造的M-估计作为目标函数,并提出了自适应确定阈值方法.仿真结果表明,使用这种方法得到的回归模型不仅具有良好的多尺度逼近特性,而且有较好的鲁棒性和较高的推广性能,具有较高的理论和应用价值.展开更多
提出了一种基于核密度估计的核偏鲁棒M-回归(kernel partial robust M-regression based on kernel density estimation,KDE-KPRM)方法。以核密度估计加权策略代替原来的M估计加权策略,利用主成分分析技术和核密度函数识别高杠杆点(输...提出了一种基于核密度估计的核偏鲁棒M-回归(kernel partial robust M-regression based on kernel density estimation,KDE-KPRM)方法。以核密度估计加权策略代替原来的M估计加权策略,利用主成分分析技术和核密度函数识别高杠杆点(输入变量空间异常点),利用残差和核密度函数识别高残差点(输出变量空间异常点),无需反复迭代便可以为样本赋予合适权重,有效地提高了建模速率。通过函数仿真和实际工业仿真,证明了所提出的方法比标准的核偏鲁棒M-回归算法有更好的鲁棒性和更高的建模效率。展开更多
Outliers accompany control engineers in their real life activity.Indus trial reality is much richer than eleme ntary linear,quadratic,Gaussian assumptions.Outliers appear due to various and varying,often unknown,reaso...Outliers accompany control engineers in their real life activity.Indus trial reality is much richer than eleme ntary linear,quadratic,Gaussian assumptions.Outliers appear due to various and varying,often unknown,reasons.They meet research interest in statistical and regression analysis and in data mining.There are a lot of interesting algorithms and approaches to outlier detection,labelling,filtering and finally interpretation.Unfortunately,their impact on control systems has not been found sufficient attention in research.Their influence is frequently unnoticed,ignored or not mentioned.This work focuses on the subject of outlier detection and labelling in the cont ext of control system performance analysis.Selec ted statistical data-driven approaches are analyzed,as t hey can be easily implemented with limited a priori knowledge.The study consists of a simulation study followed by the analysis of real control data.Differe nt generation mechanisms are Simula ted,like overlapping Gaussian processes,symmetric and asymmetric,artificially shifted points and fat-tailed distributions.Simulation observations are confronted with industrial control loops datasets.The work concludes with a practical procedure,which should help practitioners in dealing with outliers in control engineering temporal data.展开更多
文摘小波网络具有小波的多尺度特性和神经网络的自学习功能,在回归估计中得到广泛的应用,但其性能受到样本中粗差的严重影响.虽然以M-估计作为目标函数可以解决这个问题,但由于其对应的影响函数由残差绝对值决定,因此如何选择初始参数值成为一个关键问题.为此,提出回归函数的小波支持向量机鲁棒估计方法(小波支持向量回归,WSVR,Wavelet Support Vector Regression).该方法中首先提出并证明了一种新的小波支持向量机(WSVM,Wavelet Support Vector Machine),用于确定初始参数值方法,这种方法能够确定合理的网络结构和合适的初始参数值,保证含有粗差的样本点的残差绝对值较大;然后使用一种构造的M-估计作为目标函数,并提出了自适应确定阈值方法.仿真结果表明,使用这种方法得到的回归模型不仅具有良好的多尺度逼近特性,而且有较好的鲁棒性和较高的推广性能,具有较高的理论和应用价值.
文摘提出了一种基于核密度估计的核偏鲁棒M-回归(kernel partial robust M-regression based on kernel density estimation,KDE-KPRM)方法。以核密度估计加权策略代替原来的M估计加权策略,利用主成分分析技术和核密度函数识别高杠杆点(输入变量空间异常点),利用残差和核密度函数识别高残差点(输出变量空间异常点),无需反复迭代便可以为样本赋予合适权重,有效地提高了建模速率。通过函数仿真和实际工业仿真,证明了所提出的方法比标准的核偏鲁棒M-回归算法有更好的鲁棒性和更高的建模效率。
文摘Outliers accompany control engineers in their real life activity.Indus trial reality is much richer than eleme ntary linear,quadratic,Gaussian assumptions.Outliers appear due to various and varying,often unknown,reasons.They meet research interest in statistical and regression analysis and in data mining.There are a lot of interesting algorithms and approaches to outlier detection,labelling,filtering and finally interpretation.Unfortunately,their impact on control systems has not been found sufficient attention in research.Their influence is frequently unnoticed,ignored or not mentioned.This work focuses on the subject of outlier detection and labelling in the cont ext of control system performance analysis.Selec ted statistical data-driven approaches are analyzed,as t hey can be easily implemented with limited a priori knowledge.The study consists of a simulation study followed by the analysis of real control data.Differe nt generation mechanisms are Simula ted,like overlapping Gaussian processes,symmetric and asymmetric,artificially shifted points and fat-tailed distributions.Simulation observations are confronted with industrial control loops datasets.The work concludes with a practical procedure,which should help practitioners in dealing with outliers in control engineering temporal data.
