软测量技术为工业过程中重要变量及难测变量的预测提供了一个有效的解决办法。然而,由于工业过程的复杂化和高昂的数据获取成本,使得标记数据与未标记数据分布不平衡。此时,构建高性能的软测量模型成为一个挑战。针对这一问题,提出了一...软测量技术为工业过程中重要变量及难测变量的预测提供了一个有效的解决办法。然而,由于工业过程的复杂化和高昂的数据获取成本,使得标记数据与未标记数据分布不平衡。此时,构建高性能的软测量模型成为一个挑战。针对这一问题,提出了一种基于时差的多输出tri-training异构软测量方法。通过构建一种新的tri-training框架,采用多输出的高斯过程回归(multi-output Gaussian process regression,MGPR)、相关向量机(multi-output relevance vector machine,MRVM)、最小二乘支持向量机(multi-output least squares support vector machine,MLSSVM)三种模型作为基线监督回归器,使用标记数据进行训练和迭代;同时,引入时间差分(time difference,TD)改进模型的动态特性,并通过卡尔曼滤波(Kalman filtering,KF)优化模型的参数,提高其预测性能;最后通过模拟污水处理平台(benchmark simulation model 1,BSM1)和实际污水处理厂对该模型进行了验证。结果表明,与传统的软测量建模方法相比,该模型能显著提高数据分布不平衡下软测量模型的自适应性和预测性能。展开更多
In engineering application,there is only one adaptive weights estimated by most of traditional early warning radars for adaptive interference suppression in a pulse reputation interval(PRI).Therefore,if the training s...In engineering application,there is only one adaptive weights estimated by most of traditional early warning radars for adaptive interference suppression in a pulse reputation interval(PRI).Therefore,if the training samples used to calculate the weight vector does not contain the jamming,then the jamming cannot be removed by adaptive spatial filtering.If the weight vector is constantly updated in the range dimension,the training data may contain target echo signals,resulting in signal cancellation effect.To cope with the situation that the training samples are contaminated by target signal,an iterative training sample selection method based on non-homogeneous detector(NHD)is proposed in this paper for updating the weight vector in entire range dimension.The principle is presented,and the validity is proven by simulation results.展开更多
Deep neural networks(DNNs)have achieved great success in many data processing applications.However,high computational complexity and storage cost make deep learning difficult to be used on resource-constrained devices...Deep neural networks(DNNs)have achieved great success in many data processing applications.However,high computational complexity and storage cost make deep learning difficult to be used on resource-constrained devices,and it is not environmental-friendly with much power cost.In this paper,we focus on low-rank optimization for efficient deep learning techniques.In the space domain,DNNs are compressed by low rank approximation of the network parameters,which directly reduces the storage requirement with a smaller number of network parameters.In the time domain,the network parameters can be trained in a few subspaces,which enables efficient training for fast convergence.The model compression in the spatial domain is summarized into three categories as pre-train,pre-set,and compression-aware methods,respectively.With a series of integrable techniques discussed,such as sparse pruning,quantization,and entropy coding,we can ensemble them in an integration framework with lower computational complexity and storage.In addition to summary of recent technical advances,we have two findings for motivating future works.One is that the effective rank,derived from the Shannon entropy of the normalized singular values,outperforms other conventional sparse measures such as the?_1 norm for network compression.The other is a spatial and temporal balance for tensorized neural networks.For accelerating the training of tensorized neural networks,it is crucial to leverage redundancy for both model compression and subspace training.展开更多
Tri-training利用无标签数据进行分类可有效提高分类器的泛化能力,但其易将无标签数据误标,从而形成训练噪声。提出一种基于密度峰值聚类的Tri-training(Tri-training with density peaks clustering,DPC-TT)算法。密度峰值聚类通过类...Tri-training利用无标签数据进行分类可有效提高分类器的泛化能力,但其易将无标签数据误标,从而形成训练噪声。提出一种基于密度峰值聚类的Tri-training(Tri-training with density peaks clustering,DPC-TT)算法。密度峰值聚类通过类簇中心和局部密度可选出数据空间结构表现较好的样本。DPC-TT算法采用密度峰值聚类算法获取训练数据的类簇中心和样本的局部密度,对类簇中心的截断距离范围内的样本认定为空间结构表现较好,标记为核心数据,使用核心数据更新分类器,可降低迭代过程中的训练噪声,进而提高分类器的性能。实验结果表明:相比于标准Tritraining算法及其改进算法,DPC-TT算法具有更好的分类性能。展开更多
In this paper,a novel train positioning method considering satellite raw observation data was proposed,which aims to promote train positioning performance from an innovative perspective of the train satellite-based po...In this paper,a novel train positioning method considering satellite raw observation data was proposed,which aims to promote train positioning performance from an innovative perspective of the train satellite-based positioning error sources.The method focused on overcoming the abnormal observations in satellite observation data caused by railway environment rather than the positioning results.Specifically,the relative positioning experimental platform was built and the zero-baseline method was firstly employed to evaluate the carrier phase data quality,and then,GNSS combined observation models were adopted to construct the detection values,which were applied to judge abnormal-data through the dual-frequency observations.Further,ambiguity fixing optimization was investigated based on observation data selection in partly-blocked environments.The results show that the proposed method can effectively detect and address abnormal observations and improve positioning stability.Cycle slips and gross errors can be detected and identified based on dual-frequency global navigation satellite system data.After adopting the data selection strategy,the ambiguity fixing percentage was improved by 29.2%,and the standard deviation in the East,North,and Up components was enhanced by 12.7%,7.4%,and 12.5%,respectively.The proposed method can provide references for train positioning performance optimization in railway environments from the perspective of positioning error sources.展开更多
针对复杂交通场景车辆检测算法自适应能力差的问题,提出了基于Co-training半监督学习方法的车辆鲁棒检测算法.首先,针对手工标记的少量样本,分别训练基于Haar-like特征的AdaBoost分类器和基于HOG(histograms of oriented gradients)特征...针对复杂交通场景车辆检测算法自适应能力差的问题,提出了基于Co-training半监督学习方法的车辆鲁棒检测算法.首先,针对手工标记的少量样本,分别训练基于Haar-like特征的AdaBoost分类器和基于HOG(histograms of oriented gradients)特征的SVM(support vector machines)分类器,使其具有一定的识别能力;然后,基于Co-training半监督学习框架,将利用2种算法进行分类得到的新样本分别加入到对方的样本库中,增加训练样本数量,再次进行分类器的训练.由于这2类特征具有冗余性,各自检测出的正负样本包含对方漏检和误检的图像.由于样本数的增加,再次训练所得到的新分类器的鲁棒性得到了很大提高,能更加准确地检测出车辆,而且由算法对未标记样本进行分类标记,不再需要人为标记,提高了车辆检测算法的自适应能力.展开更多
针对Tri-training算法利用无标记样例时会引入噪声且限制无标记样例的利用率而导致分类性能下降的缺点,提出了AR-Tri-training(Tri-training with assistant and rich strategy)算法。提出辅助学习策略,结合富信息策略设计辅助学习器,...针对Tri-training算法利用无标记样例时会引入噪声且限制无标记样例的利用率而导致分类性能下降的缺点,提出了AR-Tri-training(Tri-training with assistant and rich strategy)算法。提出辅助学习策略,结合富信息策略设计辅助学习器,并将辅助学习器应用在Tri-training训练以及说话声识别中。实验结果表明,辅助学习器在Tri-training训练的基础上不仅降低每次迭代可能产生的误标记样例数,而且能够充分地利用无标记样例以及在验证集上的错分样例信息。从实验结果可以得出,该算法能够弥补Tri-training算法的缺点,进一步提高测试率。展开更多
文摘软测量技术为工业过程中重要变量及难测变量的预测提供了一个有效的解决办法。然而,由于工业过程的复杂化和高昂的数据获取成本,使得标记数据与未标记数据分布不平衡。此时,构建高性能的软测量模型成为一个挑战。针对这一问题,提出了一种基于时差的多输出tri-training异构软测量方法。通过构建一种新的tri-training框架,采用多输出的高斯过程回归(multi-output Gaussian process regression,MGPR)、相关向量机(multi-output relevance vector machine,MRVM)、最小二乘支持向量机(multi-output least squares support vector machine,MLSSVM)三种模型作为基线监督回归器,使用标记数据进行训练和迭代;同时,引入时间差分(time difference,TD)改进模型的动态特性,并通过卡尔曼滤波(Kalman filtering,KF)优化模型的参数,提高其预测性能;最后通过模拟污水处理平台(benchmark simulation model 1,BSM1)和实际污水处理厂对该模型进行了验证。结果表明,与传统的软测量建模方法相比,该模型能显著提高数据分布不平衡下软测量模型的自适应性和预测性能。
基金supported by the National Natural Science Foundation of China(62371049)。
文摘In engineering application,there is only one adaptive weights estimated by most of traditional early warning radars for adaptive interference suppression in a pulse reputation interval(PRI).Therefore,if the training samples used to calculate the weight vector does not contain the jamming,then the jamming cannot be removed by adaptive spatial filtering.If the weight vector is constantly updated in the range dimension,the training data may contain target echo signals,resulting in signal cancellation effect.To cope with the situation that the training samples are contaminated by target signal,an iterative training sample selection method based on non-homogeneous detector(NHD)is proposed in this paper for updating the weight vector in entire range dimension.The principle is presented,and the validity is proven by simulation results.
基金supported by the National Natural Science Foundation of China(62171088,U19A2052,62020106011)the Medico-Engineering Cooperation Funds from University of Electronic Science and Technology of China(ZYGX2021YGLH215,ZYGX2022YGRH005)。
文摘Deep neural networks(DNNs)have achieved great success in many data processing applications.However,high computational complexity and storage cost make deep learning difficult to be used on resource-constrained devices,and it is not environmental-friendly with much power cost.In this paper,we focus on low-rank optimization for efficient deep learning techniques.In the space domain,DNNs are compressed by low rank approximation of the network parameters,which directly reduces the storage requirement with a smaller number of network parameters.In the time domain,the network parameters can be trained in a few subspaces,which enables efficient training for fast convergence.The model compression in the spatial domain is summarized into three categories as pre-train,pre-set,and compression-aware methods,respectively.With a series of integrable techniques discussed,such as sparse pruning,quantization,and entropy coding,we can ensemble them in an integration framework with lower computational complexity and storage.In addition to summary of recent technical advances,we have two findings for motivating future works.One is that the effective rank,derived from the Shannon entropy of the normalized singular values,outperforms other conventional sparse measures such as the?_1 norm for network compression.The other is a spatial and temporal balance for tensorized neural networks.For accelerating the training of tensorized neural networks,it is crucial to leverage redundancy for both model compression and subspace training.
文摘Tri-training利用无标签数据进行分类可有效提高分类器的泛化能力,但其易将无标签数据误标,从而形成训练噪声。提出一种基于密度峰值聚类的Tri-training(Tri-training with density peaks clustering,DPC-TT)算法。密度峰值聚类通过类簇中心和局部密度可选出数据空间结构表现较好的样本。DPC-TT算法采用密度峰值聚类算法获取训练数据的类簇中心和样本的局部密度,对类簇中心的截断距离范围内的样本认定为空间结构表现较好,标记为核心数据,使用核心数据更新分类器,可降低迭代过程中的训练噪声,进而提高分类器的性能。实验结果表明:相比于标准Tritraining算法及其改进算法,DPC-TT算法具有更好的分类性能。
基金Project(52272339)supported by the National Natural Science Foundation of ChinaProject(2023YFB390730303)supported by the National Key Research and Development Program of China+2 种基金Project(L2023G004)supported by the Science and Technology Research and Development Program of China State Railway Group Co.,Ltd.Project(QZKFKT2023-005)supported by the State Key Laboratory of Heavy-duty and Express High-power Electric Locomotive,ChinaProject(2022JZZ05)supported by the Open Foundation of MOE Key Laboratory of Engineering Structures of Heavy Haul Railway(Central South University),China。
文摘In this paper,a novel train positioning method considering satellite raw observation data was proposed,which aims to promote train positioning performance from an innovative perspective of the train satellite-based positioning error sources.The method focused on overcoming the abnormal observations in satellite observation data caused by railway environment rather than the positioning results.Specifically,the relative positioning experimental platform was built and the zero-baseline method was firstly employed to evaluate the carrier phase data quality,and then,GNSS combined observation models were adopted to construct the detection values,which were applied to judge abnormal-data through the dual-frequency observations.Further,ambiguity fixing optimization was investigated based on observation data selection in partly-blocked environments.The results show that the proposed method can effectively detect and address abnormal observations and improve positioning stability.Cycle slips and gross errors can be detected and identified based on dual-frequency global navigation satellite system data.After adopting the data selection strategy,the ambiguity fixing percentage was improved by 29.2%,and the standard deviation in the East,North,and Up components was enhanced by 12.7%,7.4%,and 12.5%,respectively.The proposed method can provide references for train positioning performance optimization in railway environments from the perspective of positioning error sources.
文摘针对Tri-training算法利用无标记样例时会引入噪声且限制无标记样例的利用率而导致分类性能下降的缺点,提出了AR-Tri-training(Tri-training with assistant and rich strategy)算法。提出辅助学习策略,结合富信息策略设计辅助学习器,并将辅助学习器应用在Tri-training训练以及说话声识别中。实验结果表明,辅助学习器在Tri-training训练的基础上不仅降低每次迭代可能产生的误标记样例数,而且能够充分地利用无标记样例以及在验证集上的错分样例信息。从实验结果可以得出,该算法能够弥补Tri-training算法的缺点,进一步提高测试率。
