Partial label learning aims to learn a multi-class classifier,where each training example corresponds to a set of candidate labels among which only one is correct.Most studies in the label space have only focused on t...Partial label learning aims to learn a multi-class classifier,where each training example corresponds to a set of candidate labels among which only one is correct.Most studies in the label space have only focused on the difference between candidate labels and non-candidate labels.So far,however,there has been little discussion about the label correlation in the partial label learning.This paper begins with a research on the label correlation,followed by the establishment of a unified framework that integrates the label correlation,the adaptive graph,and the semantic difference maximization criterion.This work generates fresh insight into the acquisition of the learning information from the label space.Specifically,the label correlation is calculated from the candidate label set and is utilized to obtain the similarity of each pair of instances in the label space.After that,the labeling confidence for each instance is updated by the smoothness assumption that two instances should be similar outputs in the label space if they are close in the feature space.At last,an effective optimization program is utilized to solve the unified framework.Extensive experiments on artificial and real-world data sets indicate the superiority of our proposed method to state-of-art partial label learning methods.展开更多
To mitigate the deleterious effects of clutter and jammer, modern radars have adopted adaptive processing techniques such as constant false alarm rate(CFAR) detectors which are widely used to prevent clutter and noise...To mitigate the deleterious effects of clutter and jammer, modern radars have adopted adaptive processing techniques such as constant false alarm rate(CFAR) detectors which are widely used to prevent clutter and noise interference from saturating the radar’s display and preventing targets from being obscured.This paper concerns with the detection analysis of the novel version of CFAR schemes(cell-averaging generalized trimmed-mean,CATM) in the presence of additional outlying targets other than the target under research. The spurious targets as well as the tested one are assumed to be fluctuating in accordance with the χ~2-model with two-degrees of freedom. In this situation, the processor performance is enclosed by the swerling models(SWI and SWII). Between these bounds, there is an important class of target fluctuation which is known as moderately fluctuating targets. The detection of this class has many practical applications. Structure of the CATM detector is described briefly. Detection performances for optimal, CAM, CA, trimmed-mean(TM) and ordered-statistic(OS) CFAR strategies have been analyzed and compared for desired probability of false alarm and determined size of the reference window. False alarm rate performance of these processors has been evaluated for different strengths of interfering signal and the effect of correlation among the target returns on the detection and false alarm performances has also been studied. Our numerical results show that, with a proper choice of trimming parameters,the novel model CAM presents an ideal detection performance outweighing that of the Neyman-Pearson detector on condition that the tested target obeys the SWII model in its fluctuation. Although the new models CAS and CAM can be treated as special cases of the CATM algorithm, their multi-target performance is modest even it has an enhancement relative to that of the classical CAcheme. Additionally, they fail to maintain the false alarm rate constant when the operating environment is of type target multiplicity. Moreover, the non-coherent integration of M pulses ameliorates the processor performance either it operates in homogeneous or multi-target environment.展开更多
部分有序数据是同时包含有序特征与无序特征的一类数据,其广泛存在于现实生活中。传统的有序分类方法或者将所有特征都视为有序特征,或者对有序与无序特征分别进行处理,忽略了二者之间的关系,这些方法难以有效解决部分有序数据上的分类...部分有序数据是同时包含有序特征与无序特征的一类数据,其广泛存在于现实生活中。传统的有序分类方法或者将所有特征都视为有序特征,或者对有序与无序特征分别进行处理,忽略了二者之间的关系,这些方法难以有效解决部分有序数据上的分类问题。针对该问题,提出一种基于特征融合的部分有序深度森林模型,称为FFDF(feature fusion-based deep forest)。利用典型相关分析的思想,设计特征融合的贡献度计算方法,将有序特征和无序特征融合到同一特征空间,统一度量二者之间的关系。对融合的特征空间进行数据粒化,降低模型处理连续变量时的复杂性。设计融合空间下的特征矩阵输入级联森林,构建部分有序的深度森林模型。在来自UCI和WEKA的13个公共数据集上与部分单调决策树、有序分类模型、深度森林模型等六种方法进行比较实验,结果表明所提方法在准确性和平均绝对误差方面均优于对比方法;与集成模型深度森林gcForest和DF21进行了时间性能上的对比实验,结果表明所提方法在时间性能上优于对比方法。展开更多
多视图偏多标记学习主要处理同时具有多个视图和多个相关标记但标记信息不完全准确的数据.现有多视图偏多标记学习方法大多采用两阶段的方式独立进行标记消歧与多标记分类,其分类性能有待提高.本文提出了一种多视图偏多标记分类与标记...多视图偏多标记学习主要处理同时具有多个视图和多个相关标记但标记信息不完全准确的数据.现有多视图偏多标记学习方法大多采用两阶段的方式独立进行标记消歧与多标记分类,其分类性能有待提高.本文提出了一种多视图偏多标记分类与标记消歧联合学习(joint learning of multi-view partial multi-label classification and label disambiguation,JL-MVPML-LD)框架.首先,对多视图特征进行多核融合并考虑不同视图的重要性;其次,自动挖掘实例相关性和标记相关性,并利用它们来促进多视图偏多标记分类和标记消歧的联合学习;最后,采用交替迭代方法进行求解.在3个数据集上27种情况下的实验结果验证了本文方法的有效性.展开更多
为探究萌芽期大蒜挥发性物质的差异,采用电子鼻、捕集阱顶空-气质联用仪(Trap head space-gas chromatography-mass spectrometry,HS-Trap-GC-MS)结合正交偏最小二乘法判别分析(Orthogonal partial least squares discriminant analysis...为探究萌芽期大蒜挥发性物质的差异,采用电子鼻、捕集阱顶空-气质联用仪(Trap head space-gas chromatography-mass spectrometry,HS-Trap-GC-MS)结合正交偏最小二乘法判别分析(Orthogonal partial least squares discriminant analysis,OPLS-DA)、香气活度值、差异性热图、相关性分析分析大蒜萌芽在0、24、48、72、96 h挥发性物质的差异。电子鼻结合OPLS-DA建立预测模型其预测能力达96.00%。GC-MS分析表明:含硫化合物是不同萌芽期大蒜的主要共有挥发性物质,含硫化合物的相对含量随萌芽时间的延长而呈递减趋势,而种类呈现出递增趋势;二烯丙基二硫醚是样品在萌芽过程中含量降低最多的物质。二烯丙基四硫醚、烯丙硫醇是样品共有关键化合物。差异性热图分析显示:除共有物质含量差异外,硫化丙烯、己醛、叠氮二羧酸二叔丁酯、丙烯醇、6-甲基-2-庚炔、5-甲基噻二唑、2-亚乙基-1,3-二硫烷、2-丙-2-炔基磺酰基丙烷、2,5-二甲基噻吩、2,5-二甲基呋喃、1-戊烯-3-醇、1,3-二噻烷的缺失进一步加大了未萌芽和萌芽大蒜气味的差异。萌芽大蒜主要共有挥发性物质的种类随萌芽时间的延长呈现递增趋势。大蒜主要挥发性物质与电子鼻大多数传感器存在显著相关性。大蒜的气味强度会随萌芽时间的延长而逐步减弱。展开更多
基金supported by the National Natural Science Foundation of China(62176197,61806155)the National Natural Science Foundation of Shaanxi Province(2020GY-062).
文摘Partial label learning aims to learn a multi-class classifier,where each training example corresponds to a set of candidate labels among which only one is correct.Most studies in the label space have only focused on the difference between candidate labels and non-candidate labels.So far,however,there has been little discussion about the label correlation in the partial label learning.This paper begins with a research on the label correlation,followed by the establishment of a unified framework that integrates the label correlation,the adaptive graph,and the semantic difference maximization criterion.This work generates fresh insight into the acquisition of the learning information from the label space.Specifically,the label correlation is calculated from the candidate label set and is utilized to obtain the similarity of each pair of instances in the label space.After that,the labeling confidence for each instance is updated by the smoothness assumption that two instances should be similar outputs in the label space if they are close in the feature space.At last,an effective optimization program is utilized to solve the unified framework.Extensive experiments on artificial and real-world data sets indicate the superiority of our proposed method to state-of-art partial label learning methods.
文摘To mitigate the deleterious effects of clutter and jammer, modern radars have adopted adaptive processing techniques such as constant false alarm rate(CFAR) detectors which are widely used to prevent clutter and noise interference from saturating the radar’s display and preventing targets from being obscured.This paper concerns with the detection analysis of the novel version of CFAR schemes(cell-averaging generalized trimmed-mean,CATM) in the presence of additional outlying targets other than the target under research. The spurious targets as well as the tested one are assumed to be fluctuating in accordance with the χ~2-model with two-degrees of freedom. In this situation, the processor performance is enclosed by the swerling models(SWI and SWII). Between these bounds, there is an important class of target fluctuation which is known as moderately fluctuating targets. The detection of this class has many practical applications. Structure of the CATM detector is described briefly. Detection performances for optimal, CAM, CA, trimmed-mean(TM) and ordered-statistic(OS) CFAR strategies have been analyzed and compared for desired probability of false alarm and determined size of the reference window. False alarm rate performance of these processors has been evaluated for different strengths of interfering signal and the effect of correlation among the target returns on the detection and false alarm performances has also been studied. Our numerical results show that, with a proper choice of trimming parameters,the novel model CAM presents an ideal detection performance outweighing that of the Neyman-Pearson detector on condition that the tested target obeys the SWII model in its fluctuation. Although the new models CAS and CAM can be treated as special cases of the CATM algorithm, their multi-target performance is modest even it has an enhancement relative to that of the classical CAcheme. Additionally, they fail to maintain the false alarm rate constant when the operating environment is of type target multiplicity. Moreover, the non-coherent integration of M pulses ameliorates the processor performance either it operates in homogeneous or multi-target environment.
文摘部分有序数据是同时包含有序特征与无序特征的一类数据,其广泛存在于现实生活中。传统的有序分类方法或者将所有特征都视为有序特征,或者对有序与无序特征分别进行处理,忽略了二者之间的关系,这些方法难以有效解决部分有序数据上的分类问题。针对该问题,提出一种基于特征融合的部分有序深度森林模型,称为FFDF(feature fusion-based deep forest)。利用典型相关分析的思想,设计特征融合的贡献度计算方法,将有序特征和无序特征融合到同一特征空间,统一度量二者之间的关系。对融合的特征空间进行数据粒化,降低模型处理连续变量时的复杂性。设计融合空间下的特征矩阵输入级联森林,构建部分有序的深度森林模型。在来自UCI和WEKA的13个公共数据集上与部分单调决策树、有序分类模型、深度森林模型等六种方法进行比较实验,结果表明所提方法在准确性和平均绝对误差方面均优于对比方法;与集成模型深度森林gcForest和DF21进行了时间性能上的对比实验,结果表明所提方法在时间性能上优于对比方法。
文摘多视图偏多标记学习主要处理同时具有多个视图和多个相关标记但标记信息不完全准确的数据.现有多视图偏多标记学习方法大多采用两阶段的方式独立进行标记消歧与多标记分类,其分类性能有待提高.本文提出了一种多视图偏多标记分类与标记消歧联合学习(joint learning of multi-view partial multi-label classification and label disambiguation,JL-MVPML-LD)框架.首先,对多视图特征进行多核融合并考虑不同视图的重要性;其次,自动挖掘实例相关性和标记相关性,并利用它们来促进多视图偏多标记分类和标记消歧的联合学习;最后,采用交替迭代方法进行求解.在3个数据集上27种情况下的实验结果验证了本文方法的有效性.
