检索结果-南通市图书馆

False Negative Sample Detection for Graph Contrastive Learning

在线全文

学校读者我要写书评

暂无评论

Tsinghua science and Technology 2024年第2期29卷 529-542页

作者： Binbin Zhang Li Wang College of data science Taiyuan University of TechnologyJinzhong 030600China

Recently,self-supervised learning has shown great potential in Graph Neural Networks (GNNs) through contrastive learning,which aims to learn discriminative features for each node without label information. The key to graph contrastive learning is data augmentation. The anchor node regards its augmented samples as positive samples,and the rest of the samples are regarded as negative samples,some of which may be positive samples. We call these mislabeled samples as “false negative” samples,which will seriously affect the final learning effect. Since such semantically similar samples are ubiquitous in the graph,the problem of false negative samples is very significant. To address this issue,the paper proposes a novel model,False negative sample Detection for Graph Contrastive Learning (FD4GCL),which uses attribute and structure-aware to detect false negative samples. Experimental results on seven datasets show that FD4GCL outperforms the state-of-the-art baselines and even exceeds several supervised methods.

关键词： graph representation learning contrastive learning false negative sample detection

Flexible, efficient, and accurate tests for epidemics

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

science China Mathematics 2024年第8期67卷 1881-1898页

作者： Linjiajie Fang Bing-Yi Jing Shen Ling Qiyue Wang Qing Yang data science and Analytics Thrust The Hong Kong University of Science and Technology Department of Statistics and data science Southern University of Science and Technology Department of Mathematics and Statistics Dalhousie University Department of Mathematics The Hong Kong University of Science and Technology

Group testing involves discovering a small subset of distinguished subjects from a large population while efficiently reducing the total number of tests. It has been widely used for industrial testing, information technology, and biology, especially epidemic screening. Tests, in reality, are noisy for the presence of false outcomes. Some tests are accurate but time-consuming, while others are cheaper but less accurate. Exactly which test to use is constrained by various considerations, such as availability, cost, accuracy, and efficiency.In this paper, we propose flexible, efficient, and accurate tests(FEATs). FEATs are based on group testing with simple but careful designs by incorporating ideas such as close contact cliques and repeated tests. FEATs could dramatically improve the efficiency or accuracy of existing tests. For example, for accurate but slow tests, the FEAT can improve efficiency multiple times without compromising accuracy. On the other hand,for fast but inaccurate tests, the FEAT can sharply reduce the false-negative rate(FNR) and significantly increase efficiency. Theoretical justifications are provided. We point out some scenarios where the FEAT can be effectively employed.

关键词： epidemiology control blood testing testing efficiency testing accuracy

Cohomology groups of a new class of Kadison-Singer algebras

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

science China Mathematics 2024年第3期67卷 593-606页

作者： Guangyu An Xing Cheng Jun Sheng School of Mathematics and data science Shaanxi University of Science and TechnologyXi’an 710021China

Let N be a maximal discrete nest on an infinite-dimensional separable Hilbert space H,ξ=∑^(∞)_(n=1)en/2n be a separating vector for the commutant N',E_(ξ)be the projection from H onto the subspace[Cξ]spanned by the vectorξ,and Q be the projection from K=H⊕H⊕H onto the closed subspace{(η,η,η)^(T):η∈H}.Suppose that L is the projection lattice generated by the projections(E_(ξ) 0 0 0 0 0 0 0 0),{(E 0 0 0 0 0 0 0 0):E∈N},(I 0 0 0 I 0 0 0 0) and Q.We show that L is a Kadison-Singer lattice with the trivial commutant.Moreover,we prove that every n-th bounded cohomology group H~n(AlgL,B(K))with coefficients in B(K)is trivial for n≥1.

关键词： Kadison-Singer algebra Kadison-Singer lattice derivation cohomology group

THE EXTREMES OF DEPENDENT CHI-PROCESSES ATTRACTED BY THE BROWN-RESNICK PROCESS

在线全文

学校读者我要写书评

暂无评论

Acta Mathematica Scientia 2024年第2期44卷 686-701页

作者：孙俊杰谭中权 College of data science Jiaxing UniversityJiaxing314001China

Motivated by some recent works on the topic of the Brown-Resnick process, we study the functional limit theorem for normalized pointwise maxima of dependent chi-processes. It is proven that the properly normalized poi... 详细信息

关键词： chi-processes Brown-Resnick process pointwise maxima functional limit theorem

A New Class of Strong Orthogonal Arrays of Strength Three

在线全文

学校读者我要写书评

暂无评论

Journal of Systems science & Complexity 2024年第3期37卷 1233-1250页

作者： WANG Chunyan LIU Min-Qian YANG Jinyu Center for applied Statistics and School of Statistics Renmin University of ChinaBeijing 100872China School of Statistics and data science LPMC&KLMDASRNankai UniversityTianjin 300071China

Strong orthogonal arrays(SOAs) were recently introduced and studied as a class of spacefilling designs for computer experiments. To surely realize better space-filling properties, SOAs of strength three or higher are desirable. In addition, orthogonality is also an important property for designs of computer experiments, because it guarantees that the estimates of the main effects are uncorrelated. This paper first provides a systematic study on the construction of(nearly) orthogonal strength-three SOAs with better space-filling properties. The newly proposed strength-three SOAs enjoy almost the same space-filling properties of strength-four SOAs, and can accommodate much more columns than the latter. Moreover, they are(nearly) orthogonal and flexible in run sizes. The construction methods are straightforward to implement, and their theoretical supports are well established. In addition to the theoretical results, many designs are tabulated for practical needs.

关键词： Computer experiment orthogonality space-filling property strong orthogonal array

Covariate-Assisted Matrix Completion with Multiple Structural Breaks

在线全文

学校读者我要写书评

暂无评论

Journal of Systems science & Complexity 2024年第2期37卷 692-728页

作者： MENG Jing FENG Long ZOU Changliang WANG Zhaojun School of Statistics and data science KLMDASRLEBPSand LPMCNankai UniversityTianjin 300071China

In matrix completion,additional covariates often provide valuable information for completing the unobserved entries of a high-dimensional low-rank matrix A.In this paper,the authors consider the matrix recovery problem when there are multiple structural breaks in the coefficient matrix β under the column-space-decomposition model A=Xβ+B.A cumulative sum(CUSUM)statistic is constructed based on the penalized estimation of β.Then the CUSUM is incorporated into the Wild Binary Segmentation(WBS)algorithm to consistently estimate the location of breaks.Consequently,a nearly-optimal recovery of A is fulfilled.Theoretical findings are further corroborated via numerical experiments and a real-data application.

关键词： Additional covariates matrix completion multiple structural breaks wild Binary Segmentation

Neural plasticity in Parkinson's disease: a neuroimaging perspective

在线全文

学校读者我要写书评

暂无评论

Neural Regeneration Research 2024年第6期19卷 1203-1205页

作者： Christina Andica Koji Kamagata Department of Radiology Juntendo University Graduate School of Medicine Faculty of Health data science Juntendo University

Parkinson’s disease(PD) is a progressive neurodegenerative disorder characterized primarily by classical motor signs of bradykinesia,rigidity,tremors,and postural instability usually manifesti ng unilate rally or at least asymmetrically.The disease involves a loss of dopaminergic neurons projecting from the substantia nigra pars compacta to the dorsal striatum,

关键词：

First Passage Density of Brownian Motion with Two-sided Piecewise Linear Boundaries

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Acta Mathematica Sinica,English Series 2024年第6期40卷 1505-1520页

作者： Zhen YU Mao Zai TIAN Center for applied Statistics School of StatisticsRenmin University of ChinaBeijing 100872P.R.China School of Mathematics and data science Changji UniversityChangji 831100P.R.China

The first passage time has many applications in fields like finance,econometrics,statistics,and biology.However,explicit formulas for the first passage density have only been obtained for a few cases.This paper derives an explicit formula for the first passage density of Brownian motion with twosided piecewise continuous boundaries which may have some points of discontinuity.Approximations are used to obtain a simplified formula for estimating the first passage density.Moreover,the results are also generalized to the case of two-sided general nonlinear boundaries.Simulations can be easily carried out with Monte Carlo method and it is demonstrated for several typical two-sided boundaries that the proposed approximation method offers a highly accurate approximation of first passage density.

关键词： Boundary non-crossing probability first density passage density two-sided piecewise continuous boundaries Brownian motion

Perpendicular-Cutdepth:Perpendicular Direction Depth Cutting data Augmentation Method

在线全文

学校读者我要写书评

暂无评论

Computers, Materials & Continua 2024年第4期79卷 927-941页

作者： Le Zou Linsong Hu Yifan Wang Zhize Wu Xiaofeng Wang Anhui Provincial Engineering Laboratory of Big data Technology Application forUrban Infrastructure School ofArtificial Intelligence and Big DataHefei UniversityHefei230601China Institute of applied Optimization School of Artificial Intelligence and Big DataHefei UniversityHefei230601China

Depth estimation is an important task in computer vision.Collecting data at scale for monocular depth estimation is challenging,as this task requires simultaneously capturing RGB images and depth information.Therefore,data augmentation is crucial for this task.Existing data augmentationmethods often employ pixel-wise transformations,whichmay inadvertently disrupt edge features.In this paper,we propose a data augmentationmethod formonocular depth estimation,which we refer to as the Perpendicular-Cutdepth method.This method involves cutting realworld depth maps along perpendicular directions and pasting them onto input images,thereby diversifying the data without compromising edge features.To validate the effectiveness of the algorithm,we compared it with existing convolutional neural network(CNN)against the current mainstream data augmentation algorithms.Additionally,to verify the algorithm’s applicability to Transformer networks,we designed an encoder-decoder network structure based on Transformer to assess the generalization of our proposed algorithm.Experimental results demonstrate that,in the field of monocular depth estimation,our proposed method,Perpendicular-Cutdepth,outperforms traditional data augmentationmethods.On the indoor dataset NYU,our method increases accuracy from0.900 to 0.907 and reduces the error rate from0.357 to 0.351.On the outdoor dataset KITTI,our method improves accuracy from 0.9638 to 0.9642 and decreases the error rate from 0.060 to 0.0598.

关键词： Perpendicular depth estimation data augmentation