时变转速下基于改进图注意力网络的轴承半监督故障诊断

新近的基于图神经网络(GNN)的轴承半监督故障诊断研究仍存在标签信息挖掘不充分和诊断场景较理想等问题。工程实际中,轴承经常运行于启停等时变转速工况,且故障标签样本的获取成本越发昂贵。针对以上挑战,该文提出时变转速下基于改进图注意力网络(GAT)的轴承半监督故障诊断新方法。基于K最近邻(KNN)算法和平滑假设(SA)设计伪标签传播策略,将标签信息沿边传播给分布相似的邻域样本,从而充分利用有限样本的标签信息。将每个振动频谱样本视为一个节点,构建基于节点级图注意力网络的半监督学习模型,通过注意力机制进一步挖掘代表性的轴承故障特征。将所提方法用于分析两组时变转速下轴承故障实验数据,结果表明所提方法能够在不超过2%的低标签率情况下,准确诊断轴承的不同故障模式,性能优于其他常用的图神经网络半监督学习方法。     

基于可变剪接紊乱的乳腺癌亚型预测分析

可变剪接与多种复杂疾病的发生、发展存在密切的联系,包括肿瘤在内的多种疾病的产生往往伴随着可变剪接的紊乱发生。现有的乳腺癌亚型分析主要是基于单个剪接异构体出发,缺少考虑亚型之间由于可变剪接紊乱造成剪接异构体在整体分布上的差异。因此该文提出了基于可变剪接紊乱的乳腺癌亚型预测方法,主要使用Jensen-Shannon(JS)散度来找寻亚型之间的可变剪接紊乱差异较大的基因,并构建反向传播(BP)神经网络模型对乳腺癌亚型进行分类。结果表明,该方法不仅能有效发现肿瘤异质性分子,在乳腺癌亚型分类方面也有较好的识别结果,其平均F1值达到0.89,且能为患者提供个性化乳腺癌亚型药物推荐。该文的研究将有效促进基于可变剪接紊乱的乳腺癌亚型研究的发展。     

射频识别技术中间件分层结构的研究和设计(英文)

简单介绍了射频识别技术(RFID)中间件及其研究现状,在研究和分析了现存食品追溯系统的局限和不足的基础上,提出设计了分层的RFID中间件体系结构,将中间件模块分成三层,并进行了深入的分析,此分层的中间件模型屏蔽了不同设备的差异性,集中管理分布的读写器设备,并按照一定的协议规则对大量的冗余数据进行过滤筛除,并从中提取语义信息;提高了各个企业、不同厂家产品的兼容性,更有利于应用于食品追溯系统中;利用发布-订阅模型实现更加有效、实时的信息共享机制。     

放线菌扫描电镜样品制备方法比较研究

目的:比较和筛选放线菌扫描电镜样品制备方法。方法:通过培养获得插片、菌饼和发酵液3种待处理样品,根据单因素试验设计原则采用双固定、单固定、磷酸缓冲液清洗、生理盐水清洗、蒸馏水清洗、丙酮脱水、乙醇脱水和直接表面镀金等8种方法对样品进行处理。结果:取样时间直接影响观察样品形态的丰富度;插片样品扫描获得的结果较为理想;不同的固定、清洗和脱水等处理过程也对样品的细节产生了影响。结论:插片样品直接干燥喷金扫描观察是一种简便、快速的方法,对于常规的放线菌观察与鉴定较为实用,也为其它微生物扫描电镜观察提供了参考。     

隐私计算研究范畴及发展趋势

随着移动互联网、云计算和大数据技术的广泛应用,电商、搜索、社交网络等服务在提供便利的同时,大数据分析使用户隐私泄露的威胁日益凸显,不同系统隐私保护策略和能力的差异性使隐私的延伸管理更加困难,同一信息的隐私保护需求随时间变化需要多种隐私保护方案的组合协同。目前已有的各类隐私保护方案大多针对单一场景,隐私缺乏定量化的定义,隐私保护的效果、隐私泄露的利益损失以及隐私保护方案融合的复杂性三者之间的关系刻画缺乏系统的计算模型。因此,在分析隐私保护研究现状的基础上,提出隐私计算的概念,对隐私计算的内涵加以界定,从隐私信息的全生命周期讨论隐私计算研究范畴,并从隐私计算模型、隐私保护场景适应的密码理论、隐私控制与抗大数据分析的隐私保护、基于信息隐藏的隐私保护以及支持高并发的隐私保护服务架构等方面展望隐私计算的发展趋势。     

A traffic balance scheme of group emotion recognition by using the service function chain

Most researches of emotion recognition focus on single person information. However, everyone's emotions will affect each other. For example, when the teacher is angry, the student's nervousness will increase. But the facial expression information of the light single is already quite large. It means that group emotion recognition will encounter a huge traffic bottleneck. Therefore, there is a vast amount of data collected by end‐devices that will be uploaded to the emotion cloud for big data analysis. Because different emotions may require different analytical methods, in the face of diverse big data, connecting different emotion clouds is a viable alternative method to extend the emotion cloud hardware. In this paper, we built a software defined networking (SDN) multi‐emotion cloud platform to connect different emotion clouds. Through the advantages of the splicing control plane and the data plane, the routing path can be changed using software. This means that the individual conditions of different students can be handled by a dedicated system via service function (SF). The load balancing effect between different emotion clouds is achieved by optimizing the SFC. In addition, we propose a SFC‐based dynamic load balancing mechanism which eliminates a large number of SFC creation processes. The simulation results show that the proposed mechanism can effectively allocate resources to different emotion clouds to achieve real‐time emotion recognition. This is the first strategy to use SFC to balance the emotion data that the teachers can change teaching policy in a timely manner in response to students' emotions.     

Modeling intrapersonal deformation subspace using GMM for palmprint identification

In this paper, an efficient model of palmprint identification is presented based on subspace density estimation using Gaussian Mixture Model （GMM）. While a few training samples are available for each person, we use intrapersonal palmprint deformations to train the global GMM instead of modeling GMMs for every class. To reduce the dimension of such variations while preserving density function of sample space, Principle Component Analysis （PCA） is used to find the principle differences and form the Intrapersonal Deformation Subspace （IDS）. After training GMM using Expectation Maximization （EM） algorithm in IDS, a maximum likelihood strategy is carried out to identify a person. Experimental results demonstrate the advantage of our method compared with traditional PCA method and single Gaussian strategy.     

宽幅SAR海上大型运动舰船目标数据集构建及识别性能分析

以TopSAR和ScanSAR成像模式为代表的宽幅合成孔径雷达(SAR)可以实现更大范围的海洋场景观测。但实现宽测绘带的同时降低了成像分辨率,因此宽幅SAR图像中的舰船目标缺乏清晰的结构特征,给大范围海上舰船目标识别带来了极大的挑战。同时,由于缺乏海上运动航母、两栖舰等大型关键目标的宽幅SAR样本数据,使得海上运动大型关键舰船目标识别更加困难。针对该问题,该文构建了宽幅SAR海上大型运动舰船目标数据集,共包含2291个大型舰船目标样本,类别划分为大型军事舰船,长度为≥250 m的大型民用舰船和长度为150～250 m的大型民用舰船。其构建过程为:首先,通过先验知识辅助获取港口区域大型军事舰船目标样本;其次,利用属性信息对OpenSARShip数据集进行长度筛选获取大型民用舰船目标样本;最后,在样本的距离-多普勒域添加二次相位误差进行运动仿真来模拟海上运动舰船目标的成像结果。此外,该文使用经典识别算法和深度学习方法对构建的数据集与运动仿真处理后的数据进行了识别性能分析,结果表明在低分辨率情况下采用SAR图像复数信息可以在一定程度上提高算法的识别率;并且运动舰船目标的散焦问题对识别准确率具有较大影响。      

基于视点差异和多分类器的三维模型分类

基于视图的3维模型分类方法与深度学习融合能有效提升模型分类的准确率。但目前的方法将相同类别的3维模型所有视点上的视图归为一类,忽略了不同视点上的视图差异,导致分类器很难学习到一个合理的分类面。为解决这一问题,该文提出一个基于深度神经网络的3维模型分类方法。该方法在3维模型的周围均匀设置多个视点组,为每个视点组训练1个视图分类器,充分挖掘不同视点组下的3维模型深度信息。这些分类器共享1个特征提取网络,但却有各自的分类网络。为了使提取的视图特征具有区分性,在特征提取网络中加入注意力机制;为了对非本视点组的视图建模,在分类网络中增加了附加类。在分类阶段首先提出一个视图选择策略,从大量视图中选择少量视图用于分类,以提高分类效率。然后提出一个分类策略通过分类视图实现可靠的3维模型分类。在ModelNet10和ModelNet40上的实验结果表明,该方法在仅用3张视图的情况下分类准确率高达93.6%和91.0%。     

Bayesian网中概率参数学习方法

Bayesian网已经成为AI领域的研究热点,并在现代专家系统、诊断系统及决策支持系统中发挥着至关重要的作用.Bayesian网的研究主要集中在三个方面:知识表示、学习与推理.概率知识是Bayesian网坚实的数学基础,从数据中学习分布参数使得Bayesian网逐步走向现实应用.本文介绍和比较了概率参数学习的各种常用方法,并探求了它们在不同应用背景下的优缺点.基于经典统计学的方法理论成熟,计算简单,但它只利用了实例数据集合所提供的信息,无法加入专家知识,对实例数据的依赖性大;基于Bayesian有机结合了两类信息,对实例数据的依赖性降低,学习结果更加准确.参数学习是Bayesian网学习的基础,是Bayesian网结构学习必不可少的部分.     

Evaluation of relevance feedback schemes in content-based in retrieval systems

Multimedia content modeling, i.e., identification of semantically meaningful entities, is an arduous task mainly due to the fact that (a) humans perceive the content using high-level concepts and (b) the subjectivity of human perception, which often interprets the same content in a different way at different times. For this reason, an efficient content management system has to be adapted to current user's information needs and preferences through an on-line learning strategy based on users’ interaction. One adaptive learning strategy is relevance feedback, originally developed in traditional text-based information retrieval systems. In this way, the user interacts with the system to provide information about the relevance of the content, which is then fed back to the system to update its performance. In this paper, we evaluate and investigate three main types of relevance feedback algorithms; the Euclidean, the query point movements and the correlation-based approaches. In the first case, we examine heuristic and optimal techniques which are based either on the weighted or the generalized Euclidean distance. In the second case, we survey single and multipoint query movement schemes. As far as the third type is concerned, two different ways for parametrizing the normalized cross-correlation similarity metric are proposed. The first scales only the elements of the query feature vector and called query-scaling strategy, while the second scales both the query and the selected samples (query-sample scaling strategy). All the examined algorithms are evaluated using both subjective and objective criteria. Subjective evaluation is performed by depicting the best retrieved images as response of the system to a user's query. Instead, objective evaluation is obtained using standard criteria, such as the precision–recall curve and the average normalized modified retrieval rank (ANMRR). Furthermore, a newly objective criterion, called average normalized similarity metric distance is introduced which exploits the difference among the actual and ideal similarity measure among all best retrievals. Discussions and comparisons of all the aforementioned relevance feedback algorithms are presented.     

有噪声标注情况下的中医舌色分类方法

舌色是中医(TCM)望诊最关注的诊察特征之一,自动准确的舌色分类是舌诊客观化研究的重要内容.由于不同类别舌色之间的视觉界限存在模糊性以及医生标注者的主观性等,标注的舌象数据中常含有噪声,影响舌色分类模型的训练.为此,该文提出一种有噪声标注情况下的中医舌色分类方法:首先,提出一种两阶段的数据清洗方法,对含有噪声的标注样本...     

The Design and Implementation of Embedded Security CPU Based on Multi-strategy

Control flow monitoring,information flow tracking and memory monitoring are the three main solutions to enhance the security of embedded system at the hardware architecture level.However,most of the current studies about the security of embedded system consider the above solutions in separate dimensions rather than a combined effort.We start from the operation model at the instruction level,and propose a security multi-strategy which combines information flow tracking and memory monitoring by studying the security operating mechanism of embedded system.As a hardware approach this strategy extends the embedded processor architecture with additional security defense control.The experimental results show this multi-strategy is more effective and can detect more malicious attacks than a single solution.The effectiveness of our proposed security multi-strategy has been verified in a Field programmable gate array (FPGA) prototype platform based on a customized Leon3 microprocessor.     

Indoor localization via -graph regularized semi-supervised manifold learning

In this paper,a new l1-graph regularized semi-supervised manifold learning(LRSML) method is proposed for indoor localization.Due to noise corruption and non-linearity of received signal strength(RSS),traditional approaches always fail to deliver accurate positioning results.The l1-graph is constructed by sparse representation of each sample with respect to remaining samples.Noise factor is considered in the construction process of l1-graph,leading to more robustness compared to traditional k-nearest-neighbor graph(KNN-graph).The KNN-graph construction is supervised,while the l1-graph is assumed to be unsupervised without harnessing any data label information and uncovers the underlying sparse relationship of each data.Combining KNN-graph and l1-graph,both labeled and unlabeled information are utilized,so the LRSML method has the potential to convey more discriminative information compared to conventional methods.To overcome the non-linearity of RSS,kernel-based manifold learning method(K-LRSML) is employed through mapping the original signal data to a higher dimension Hilbert space.The efficiency and superiority of LRSML over current state of art methods are verified with extensive experiments on real data.     

基于群体决策的多用户协同交互式遗传算法

采用交互式遗传算法求解大数据信息检索问题时,为实现偏好信息的提取和优化,单用户需完成较多数量的人-机交互操作,由此易产生用户疲劳、算法搜索效率低的难题。对此,该文在算法中引入多用户并行策略,通过群体决策优势,提高样本利用效率。首先,根据优化目标性质确定共性化协同或个性化协同类型,基于用户浏览行为计算用户相似度和个体相似度。然后,通过共享偏好相似用户的偏好相似个体预测个体区间适应值。基于个体表现型相似度聚类,提出大规模种群个体“区间数-区间数”适应值赋值策略。最后,依据子代种群个体与父代种群最优个体的相似性,推荐用户最佳评价个体。将所提方法应用于装饰性墙壁纸选型问题,并与已有典型方法比较。结果表明,所提方法在推荐个体质量、减轻用户疲劳、提高搜索效率等方面均具有优越性。     

模糊近邻密度聚类与重采样的迁移学习算法

传统机器学习要求训练样本和测试样本具有相同分布的假设在实际应用中难以满足,为解决这种问题,迁移学习的研究近年来逐渐兴起。其中,基于聚类分析与重采样的迁移学习框架不需要直接估计域分布,且能够修正不同类型的域间差异,但其所采用的聚类算法对参数选择的鲁棒性及不同分布数据的适应性较差,并不能很好地适用于挖掘数据结构信息。为此,该文提出一种基于模糊近邻密度聚类与重采样的迁移学习算法。该方法对不同分布形状和密度的数据具有较好的鲁棒性并能够发现更多的近邻结构信息,能够从源域中迁移更多的有用知识用于目标域的学习。在公共数据集上的实验结果表明所提出的迁移学习方法具有更好的性能。      

基于邻域最小生成树的半监督极化SAR图像分类方法

该文针对极化SAR图像分类中只有少量标记样本的问题,提出了一种基于邻域最小生成树的半监督极化SAR图像分类方法。该方法针对极化SAR图像以像素为分类对象的特点,结合自训练方法的思想,利用极化SAR图像像素点的空间信息,提出了基于邻域最小生成树辅助学习的样本选择策略,增加自训练过程中被选择无标记样本的可靠性,扩充标记样本数量,训练更好的分类器。最终用训练好的分类器对极化SAR图像进行测试。对3组真实的极化SAR图像进行测试,实验结果表明,该方法在只有少量标记样本的情况下能获得满意的分类结果,且分类正确率明显优于传统的分类算法。      

基于一致性预测器的卷烟牌号的鉴别研究

目前对于卷烟牌号的鉴别多应用一些传统分类算法,这些传统算法用于归纳一个通用规则的训练样本数据较少,造成分类模型的准确度较低,且预测结果没有置信度衡量,在高风险领域的应用不足。针对传统分类算法的局限性,提出了基于转导推理的一致性预测算法。通过探索待测数据和样本序列之间的内在联系,运用Kolmogorov算法的随机性理论建立一种置信度机制,并应用随机性检测函数对置信度进行估算,这样可以很好地对烟叶和成品卷烟进行定性判别和分类。     

基于深度学习的云参量反演方法研究

有效的云检测与云相态判识对于农业、气候及人类生活具有重要意义, 而这些数据的获取离不开卫星遥感。 卫星遥感数据在当今社会的生产和生活中都扮演着至关重要的角色, 众多领域的发展都离不开卫星遥感数据的支持。 随着高精度传感器的发展, 传统研究方法无法满足大规模、高维度数据的高效挖掘与处理, 因此深度学习技术在遥感 领域得到了快速的发展。基于深度学习技术提出了一种结合多波段遥感影像的云检测及云相态判识的方法。该方法 采用 MODIS 云产品影像作为样本, 将不同波段信息作为特征值, 分别建立针对云检测与云相态判识研究任务的多个 数据库, 并采用 DeepLab V3+ 模型进行训练并预测, 从而完成高精度的云检测及云相态判识任务。与传统方法相比, 该方法高效便捷、特征提取能力较强, 将多波段作为特征值输入模型进行预测时, 该方法展现了良好的结果。     

Comparison of single-sensor current control in the DC link forthree-phase voltage-source PWM converters

This paper presents a technique for reconstructing converter line currents using the information from a single current sensor in the DC link of a converter and comparative evaluation of the performance of single sensor control techniques in the DC link for voltage-source pulsewidth modulation (V-S PWM) converters. When 3φ input currents cannot be reconstructed, three methods to acquire the DC-link current are compared. Two of them are methods of modifying the switching state (I, II), and another is a method of using the predictive state observer. Also, compensation of sampling delay, and a simultaneous sample value of input currents in the center of a switching period are included. Suitable criteria for the comparison are identified, and the differences in the performance of these methods are investigated through experimental results for a typical V-S PWM converter rated at 10 kVA