A novel stability computation approach for tactical missile autopilots is detailed. The limi- tations of traditional stability margins are exhibited. Then the vector margin is introduced and com- pared with sensitivit...A novel stability computation approach for tactical missile autopilots is detailed. The limi- tations of traditional stability margins are exhibited. Then the vector margin is introduced and com- pared with sensitivity function to show their essential relationship. The longitudinal three-loop auto- pilot for tactical missiles is presented and used as the baseline for all the available linear autopilots. Ten linear autopilot topologies using all the measurable feedback components are given with the iden- tical closed-loop characteristic equation and time-domain step response. However, the stability of the ten autopilots differs when considering the actuator dynamics, which limits their application. Then vector margin method is adopted to compute and evaluate the stability of all available autopi- lots. The analysis and computation results show that the vector margin method could better evaluate autopilot stability.展开更多
最小二乘支持向量机(Least Squares Support Vector Machine,LSSVM)通过求解一个线性等式方程组来提高支持向量机(Support Vector Machine,SVM)的运算速度。但是,LSSVM没有考虑间隔分布对于LSSVM模型的影响,导致其精度较低。为了增强LS...最小二乘支持向量机(Least Squares Support Vector Machine,LSSVM)通过求解一个线性等式方程组来提高支持向量机(Support Vector Machine,SVM)的运算速度。但是,LSSVM没有考虑间隔分布对于LSSVM模型的影响,导致其精度较低。为了增强LSSVM模型的泛化性能,提高其分类能力,提出一种具有间隔分布优化的最小二乘支持向量机(LSSVM with margin distribution optimization,MLSSVM)。首先,重新定义间隔均值和间隔方差,深入挖掘数据的间隔分布信息,增强模型的泛化性能;其次,引入权重线性损失,进一步优化了间隔均值,提升模型的分类精度;然后,分析目标函数,剔除冗余项,进一步优化间隔方差;最后,保留LSSVM的求解机制,保障模型的计算效率。实验表明,新提出的分类模型具有良好的泛化性能和运行时间。展开更多
针对最小二乘双参数间隔支持向量机(LSTPMSVM)对噪声敏感且在分类过程中易受异常值影响的问题,提出了一种鲁棒的模糊最小二乘双参数间隔支持向量机算法(RFLSTPMSVM).该算法利用松弛变量的2范数使得优化问题具有强凸性,再根据隶属度为每...针对最小二乘双参数间隔支持向量机(LSTPMSVM)对噪声敏感且在分类过程中易受异常值影响的问题,提出了一种鲁棒的模糊最小二乘双参数间隔支持向量机算法(RFLSTPMSVM).该算法利用松弛变量的2范数使得优化问题具有强凸性,再根据隶属度为每个样本分配相应的权重,有效降低异常值带来的影响.同时,在目标函数中引入K-近邻加权,考虑样本之间的局部信息,提高模型的分类准确率.此外,通过求解简单的线性方程组来优化该算法,而不是求解二次规划问题,使模型具有较快的计算速度.在UCI(university of California irvine)数据集上对该算法进行性能评估,并与TWSVM、LSTSVM、LSTPMSVM和ULSTPMSVM 4种算法进行比较.数值实验结果表明,该算法具有更好的泛化性能.展开更多
加权线性损失孪生支持向量机(weighted linear loss twin support vector machine,WLTSVM)是针对大规模问题而构建的支持向量机(support vector machine,SVM)模型,其表现出较好的泛化性能。与SVM类似,WLTSVM中并未考虑样本集整体的间隔...加权线性损失孪生支持向量机(weighted linear loss twin support vector machine,WLTSVM)是针对大规模问题而构建的支持向量机(support vector machine,SVM)模型,其表现出较好的泛化性能。与SVM类似,WLTSVM中并未考虑样本集整体的间隔分布,而间隔分布已经被证明对泛化性能起着至关重要的作用。因此,为了取得更优性能,提出了一个加权线性损失孪生大间隔分布机(weighted linear loss twin large margin distribution machine,WLTLDM)。WLTLDM以一对非平行超平面模型为基础,构建了其对应的间隔均值和方差来优化间隔分布,并采用加权线性损失函数来度量类内样本的损失,使用平方损失函数来度量类间样本的损失。在真实数据集上的实验结果表明,WLTLDM是一个有效的间隔分布模型,其性能显著优于其他基准模型。展开更多
As a set of supervised pattern recognition methods, support vector machines (SVMs) have been successfully applied to functional magnetic resonance imaging (fMRI) field, but few studies have focused on visualizing disc...As a set of supervised pattern recognition methods, support vector machines (SVMs) have been successfully applied to functional magnetic resonance imaging (fMRI) field, but few studies have focused on visualizing discriminative regions of whole brain between different cognitive tasks dynamically. This paper presents a SVM-based method for visualizing dynamically discriminative activation of whole-brain voxels between two kinds of tasks without any contrast. Our method provides a series of dynamic spatial discrimination maps (DSDMs), representing the temporal evolution of discriminative brain activation during a duty cycle and describing how the discriminating information changes over the duty cycle. The proposed method was applied to investigate discriminative brain functional activations of whole brain voxels dynamically based on a hand-motor task experiment. A set of DSDMs between left hand movement and right hand movement were reached. Our results demonstrated not only where but also when the discriminative activations of whole brain voxels occurred between left hand movement and right hand movement during one duty cycle.展开更多
现有的一分类支持向量机算法基于优化最小间隔的思想,只考虑了样本靠近空间原点一侧的噪声,对噪声信息较为敏感。针对该问题,通过优化间隔分布思想,同时考虑样本靠近空间原点和远离空间原点两侧的噪声,提高一分类支持向量机算法的抗噪...现有的一分类支持向量机算法基于优化最小间隔的思想,只考虑了样本靠近空间原点一侧的噪声,对噪声信息较为敏感。针对该问题,通过优化间隔分布思想,同时考虑样本靠近空间原点和远离空间原点两侧的噪声,提高一分类支持向量机算法的抗噪声能力。为此,提出了一种基于最优间隔分布的一分类学习方法(one-class optimal margin distribution machine,OCODM),该方法通过最大化间隔的均值和最小化间隔方差的方式来优化间隔分布。实验结果表明,相比于现有的一分类支持向量机算法,该方法具有更好的鲁棒性,是现有一分类支持向量机方法的有益补充,能够增强现有方法的抗噪声能力。展开更多
针对非平行支持向量机(NonParallel Support Vector Machine,NPSVM)对噪声敏感和忽略了数据分布结构的问题,提出了一种具有间隔分布的抗噪声非平行支持向量机(Anti-Noise NPSVM with Margin Distribution, MDANPSVM)分类模型.在MD-ANPSV...针对非平行支持向量机(NonParallel Support Vector Machine,NPSVM)对噪声敏感和忽略了数据分布结构的问题,提出了一种具有间隔分布的抗噪声非平行支持向量机(Anti-Noise NPSVM with Margin Distribution, MDANPSVM)分类模型.在MD-ANPSVM模型中,每个优化问题同时最小化两类样本的基于L1范数的绝对损失和改进的铰链损失,这可以保证模型的稳定性,减小噪声和异常值的影响.此外,在MD-ANPSVM模型中,采用一阶和二阶统计量来描述训练数据的间隔分布信息,并试图同时最大化间隔均值和最小化间隔方差,这进一步提高了模型的泛化性能.最终,我们在不同的数据集上进行了对比实验.实验结果显示,MD-ANPSVM模型具有较强的泛化能力和强鲁棒性.展开更多
对于二类目标特征选择问题,首先讨论了特征空间的线性可分性问题,并给出了其判别条件;其次,通过借鉴支撑矢量机原理,分析了特征可分性判据的基本性质;最后,依据各特征对分类间隔的贡献大小定义了特征有效率,并以此进行特征选择和特征空...对于二类目标特征选择问题,首先讨论了特征空间的线性可分性问题,并给出了其判别条件;其次,通过借鉴支撑矢量机原理,分析了特征可分性判据的基本性质;最后,依据各特征对分类间隔的贡献大小定义了特征有效率,并以此进行特征选择和特征空间降维.实测数据与网络公开UCI(University of california,Irvine)数据库的实验结果表明,与经典的Relief特征选择算法相比,该算法在识别性能和推广能力上明显有所提高.展开更多
基金Supported by the National Natural Science Foundation of China(61172182)
文摘A novel stability computation approach for tactical missile autopilots is detailed. The limi- tations of traditional stability margins are exhibited. Then the vector margin is introduced and com- pared with sensitivity function to show their essential relationship. The longitudinal three-loop auto- pilot for tactical missiles is presented and used as the baseline for all the available linear autopilots. Ten linear autopilot topologies using all the measurable feedback components are given with the iden- tical closed-loop characteristic equation and time-domain step response. However, the stability of the ten autopilots differs when considering the actuator dynamics, which limits their application. Then vector margin method is adopted to compute and evaluate the stability of all available autopi- lots. The analysis and computation results show that the vector margin method could better evaluate autopilot stability.
文摘最小二乘支持向量机(Least Squares Support Vector Machine,LSSVM)通过求解一个线性等式方程组来提高支持向量机(Support Vector Machine,SVM)的运算速度。但是,LSSVM没有考虑间隔分布对于LSSVM模型的影响,导致其精度较低。为了增强LSSVM模型的泛化性能,提高其分类能力,提出一种具有间隔分布优化的最小二乘支持向量机(LSSVM with margin distribution optimization,MLSSVM)。首先,重新定义间隔均值和间隔方差,深入挖掘数据的间隔分布信息,增强模型的泛化性能;其次,引入权重线性损失,进一步优化了间隔均值,提升模型的分类精度;然后,分析目标函数,剔除冗余项,进一步优化间隔方差;最后,保留LSSVM的求解机制,保障模型的计算效率。实验表明,新提出的分类模型具有良好的泛化性能和运行时间。
文摘针对最小二乘双参数间隔支持向量机(LSTPMSVM)对噪声敏感且在分类过程中易受异常值影响的问题,提出了一种鲁棒的模糊最小二乘双参数间隔支持向量机算法(RFLSTPMSVM).该算法利用松弛变量的2范数使得优化问题具有强凸性,再根据隶属度为每个样本分配相应的权重,有效降低异常值带来的影响.同时,在目标函数中引入K-近邻加权,考虑样本之间的局部信息,提高模型的分类准确率.此外,通过求解简单的线性方程组来优化该算法,而不是求解二次规划问题,使模型具有较快的计算速度.在UCI(university of California irvine)数据集上对该算法进行性能评估,并与TWSVM、LSTSVM、LSTPMSVM和ULSTPMSVM 4种算法进行比较.数值实验结果表明,该算法具有更好的泛化性能.
文摘加权线性损失孪生支持向量机(weighted linear loss twin support vector machine,WLTSVM)是针对大规模问题而构建的支持向量机(support vector machine,SVM)模型,其表现出较好的泛化性能。与SVM类似,WLTSVM中并未考虑样本集整体的间隔分布,而间隔分布已经被证明对泛化性能起着至关重要的作用。因此,为了取得更优性能,提出了一个加权线性损失孪生大间隔分布机(weighted linear loss twin large margin distribution machine,WLTLDM)。WLTLDM以一对非平行超平面模型为基础,构建了其对应的间隔均值和方差来优化间隔分布,并采用加权线性损失函数来度量类内样本的损失,使用平方损失函数来度量类间样本的损失。在真实数据集上的实验结果表明,WLTLDM是一个有效的间隔分布模型,其性能显著优于其他基准模型。
文摘As a set of supervised pattern recognition methods, support vector machines (SVMs) have been successfully applied to functional magnetic resonance imaging (fMRI) field, but few studies have focused on visualizing discriminative regions of whole brain between different cognitive tasks dynamically. This paper presents a SVM-based method for visualizing dynamically discriminative activation of whole-brain voxels between two kinds of tasks without any contrast. Our method provides a series of dynamic spatial discrimination maps (DSDMs), representing the temporal evolution of discriminative brain activation during a duty cycle and describing how the discriminating information changes over the duty cycle. The proposed method was applied to investigate discriminative brain functional activations of whole brain voxels dynamically based on a hand-motor task experiment. A set of DSDMs between left hand movement and right hand movement were reached. Our results demonstrated not only where but also when the discriminative activations of whole brain voxels occurred between left hand movement and right hand movement during one duty cycle.
文摘现有的一分类支持向量机算法基于优化最小间隔的思想,只考虑了样本靠近空间原点一侧的噪声,对噪声信息较为敏感。针对该问题,通过优化间隔分布思想,同时考虑样本靠近空间原点和远离空间原点两侧的噪声,提高一分类支持向量机算法的抗噪声能力。为此,提出了一种基于最优间隔分布的一分类学习方法(one-class optimal margin distribution machine,OCODM),该方法通过最大化间隔的均值和最小化间隔方差的方式来优化间隔分布。实验结果表明,相比于现有的一分类支持向量机算法,该方法具有更好的鲁棒性,是现有一分类支持向量机方法的有益补充,能够增强现有方法的抗噪声能力。
文摘针对非平行支持向量机(NonParallel Support Vector Machine,NPSVM)对噪声敏感和忽略了数据分布结构的问题,提出了一种具有间隔分布的抗噪声非平行支持向量机(Anti-Noise NPSVM with Margin Distribution, MDANPSVM)分类模型.在MD-ANPSVM模型中,每个优化问题同时最小化两类样本的基于L1范数的绝对损失和改进的铰链损失,这可以保证模型的稳定性,减小噪声和异常值的影响.此外,在MD-ANPSVM模型中,采用一阶和二阶统计量来描述训练数据的间隔分布信息,并试图同时最大化间隔均值和最小化间隔方差,这进一步提高了模型的泛化性能.最终,我们在不同的数据集上进行了对比实验.实验结果显示,MD-ANPSVM模型具有较强的泛化能力和强鲁棒性.
文摘对于二类目标特征选择问题,首先讨论了特征空间的线性可分性问题,并给出了其判别条件;其次,通过借鉴支撑矢量机原理,分析了特征可分性判据的基本性质;最后,依据各特征对分类间隔的贡献大小定义了特征有效率,并以此进行特征选择和特征空间降维.实测数据与网络公开UCI(University of california,Irvine)数据库的实验结果表明,与经典的Relief特征选择算法相比,该算法在识别性能和推广能力上明显有所提高.