检索结果-内蒙古大学图书馆

International Conference on Multimedia Communications

作者： Zhang, Xiayi Liu, Fuqiang Li, Zhipeng Key Laboratory of Embedded System and Service Computing Tongji University Shanghai China

ISBN: (纸本)9780769541365

Statistical background subtraction has proved to be a robust and effective approach for segmenting and extracting objects without any prior information of the foreground objects. This paper presents two contributions on this topic. The first contribution of this paper proposes a novel approach which introduces the motion mask into the Gaussian Mixture Models to reduce the errors of classical GMMs, which always classifies the moving objects as background incorrectly, and affects the accuracy of the steps followed by, when the objects are still in long periods. The second contribution regards the connected component labeling based on the contour tracking algorithm. Experimental results validate the effectiveness of the proposed approach. © 2010 IEEE.

关键词： Object recognition

来源：评论

学校读者我要写书评

暂无评论

A credible individual behavior profiling method for online payment fraud detection 21

A credible individual behavior profiling method for online p...

引用

4th International Conference on Data Storage and Data Engineering, DSDE 2021

作者： Cui, Jipeng Yan, Chungang Wang, Cheng Key Laboratory of the Ministry of Education for Embedded System and Service Computing Department of Computer Science and Engineering Tongji University China

ISBN: (纸本)9781450389303

Online payment fraud detection relies on a credible characterization of individual behaviors. Existing individual behavior profiling methods have three problems in credibility. Firstly, they solely regard users as individuals to model and fail to credibly characterize the whole behavior patterns of transactions. Secondly, they cannot make full use of the general information among similar individuals due to the heterogeneity of transaction attributes. Thirdly, they are not capable of utilizing the label information of transactions credibly. Faced with these challenges, we propose a credible method for individual behavior profiling that consists of two steps. The first step is the construction of the credible individual behavior profiling framework. To begin with, the concept of individuals is generalized and the credible transaction description is given. Afterwards, based on the co-occurrence information of transaction attributes, the credible behavior condition is defined to set a limit to individual behavior patterns. The second step is the implementation of the framework. For this purpose, the original fraud detection problem is transformed into a pseudo-recommender system problem with a co-occurrence mapping process. In order to guarantee the credible behavior constraint, a credible recommendation algorithm is designed to collaboratively restore both the ranking and rating information of the pseudo-rating matrices. Besides, an embedding based approach is adopted to parameterize the proposed model. Experiments on an online payment transaction dataset demonstrate the effectiveness of our proposed method. © 2021 ACM.

关键词： Crime

来源：评论

学校读者我要写书评

暂无评论

Efficient Rate Control Algorithm for Hierarchical Video Coding

引用

Journal of Donghua University(English Edition) 2012年第3期29卷 222-226页

作者：陆千里岑锋许维胜朱芳来 The Key Laboratory of Embedded System and Service Computing Ministry of EducationTongji University Department of Control Science&Engineering Tongji University

For rate control (RC) of hierarchical structure coding, an independent rate-quantization (R-Q) model was proposed based on mean absolute differences (MADs) in different temporal levels (TLs). In the proposed R-Q model, a novel MAD model was developed according to the hierarchical structure. The experimental results demonstrate that the proposed algorithm provides better performance, in terms of average peak signal-to-noise ratio (PSNR) and quality smoothness, than the H.264 reference model, JM14.2, under various sequences.

关键词： rate-quantization (R-Q) model rate control(RC) video coding peak signal-to-noise ratio (PSNR) hierarchical structure

来源：评论

学校读者我要写书评

暂无评论

Robust Learning to Rank Based on Portfolio Theory and AMOSA Algorithm

引用

IEEE Transactions on systems, Man, and Cybernetics: systems 2017年第6期47卷 1007-1018页

作者： Li, Jinzhong Liu, Guanjun Yan, Chungang Jiang, Changjun Key Laboratory of the Ministry of Education for Embedded System and Service Computing Department of Computer Science and Technology Tongji University Shanghai201804 China Key Laboratory of Watershed Ecology and Geographical Environment Monitoring NASG Department of Computer Science and Technology Jinggangshan University Ji'an343009 China

Effectiveness is the most important factor considered in the ranking models yielded by algorithms of learning to rank (LTR). Most of the related ranking models only focus on improving the average effectiveness but ignore robustness. When a ranking model ignores robustness, the effectiveness for many queries is possibly very poor although the average effectiveness for all queries is relatively high. Therefore, Wang et al. first consider robustness in their ranking models. However, the robustness formula defined by Wang et al. cannot characterize those queries whose effectiveness are hurt seriously in comparison with the baseline model. In order to overcome this shortcoming, we propose a novel formula of characterizing robustness based on portfolio theory, and construct a multiobjective optimization model of the robust LTR in which the formula is used. Based on this model, we propose an approach of risk-sensitive and robust LTR, named as R2Rank, which is based on the framework of archived multiobjective simulated annealing algorithm and the idea of preference ranking organization method for enrichment evaluation. The experimental results show that the ranking models produced by our proposed R2Rank approach are better in both effectiveness and robustness than those produced by three state-of-the-art LTR approaches. © 2016 IEEE.

关键词： Simulated annealing

来源：评论

学校读者我要写书评

暂无评论

Neighborhood Extended Dynamic Graph Neural Network 2022

Neighborhood Extended Dynamic Graph Neural Network

引用

14th International Conference on Machine Learning and Computing, ICMLC 2022

作者： Yu, Da Wang, Junli Jiang, Changjun Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University China

ISBN: (纸本)9781450395700

Representation learning on dynamic graphs has drawn much attention due to its ability to learn hidden relationships as well as capture temporal patterns in graphs. It can be applied to represent a broad spectrum of graph-based data like social networks and further use the learned representations to solve the downstream tasks including link prediction and edge classification. Although some approaches have been proposed for dynamic graphs in recent years, most of them paid little attention to the evolution of the entire graph topology, leading to the lack of global information of what happened in nodes' neighborhoods during their update. We propose NEDGNN, a novel Neighborhood Extended Dynamic Graph Neural Network on dynamic graphs represented as sequences of time-stamped events. We introduce a temporal attention propagation module to generate messages for n-hop neighbors through a self-attention mechanism, which can help disseminate information among nodes' n-hop neighbors. Besides, a FIFO message box module is also applied to gain some time efficiency. Due to the introduction of these modules, NEDGNN outperforms many state-of-the-art baselines in several tasks. We also perform a detailed ablation study to test the effectiveness and time cost of each module. © 2022 ACM.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

A metadata management strategy based on event-classification in intelligent transportation system 15th

A metadata management strategy based on event-classification...

引用

15th International Conference on Algorithms and Architectures for Parallel Processing, ICA3PP 2015

作者： Su, Yayun Zhang, Yaying Key Laboratory of Embedded System and Service Computing Tongji University Shanghai200092 China

ISBN: (纸本)9783319271187

With the explosive growth of data information, the object-oriented storage system has been widely used. This paper proposed a metadata management strategy based on Distributed File system-Ceph in terms of event classification, taking advantage of the characteristics of data in urban traffic system. The large amount of data with a wide variety of sources and data types was first classified by machine learning, and a classification model was established. Then, improvements on load balancing were made to the existing Ceph Load Balancing Strategy of metadata partition. The metadata partitioning is to assign and migrate metadata obtained from the event classification model to the target server chosen by the fuzzy optimum method. Experimental results show that the proposed load balancing strategy based on event classification can not only make the overall load of the metadata servers in a relatively stable state but also make the migration times less than that of other algorithms. The extra overhead of the system is also reduced. © Springer International Publishing Switzerland 2015.

关键词： Metadata

来源：评论

学校读者我要写书评

暂无评论

A novel storing and accessing method of traffic incident video based on spatial-temporal analysis 15th

A novel storing and accessing method of traffic incident vid...

引用

15th International Conference on Algorithms and Architectures for Parallel Processing, ICA3PP 2015

作者： Zhang, Yaying Zhu, Yinyin Key Laboratory of Embedded System and Service Computing Tongji University Shanghai201804 China

ISBN: (纸本)9783319271217

Hadoop Distributed File system (HDFS) is a reliable and scalable data storage solution. However, it has great weakness in storage of the numerous small files. A merging method of small video files containing traffic incidents is proposed to improve the HDFS storage efficiency of small files. As traffic incident videos can be classified in terms of time and the crossroad where the incident happens, the proposed method merges video files together by time and region (usually adjacent crossroads). Indexing mechanism has been improved in later searching for small video files. The whole HDFS file block related to specific incidents will be read out to local cache. The experimental results show that when accessing for traffic incidents by region in certain period, the average search time and the memory load of HDFS NameNode will be effectively reduced. © Springer International Publishing Switzerland 2015.

关键词： Storage efficiency

来源：评论

学校读者我要写书评

暂无评论

On complementary effect of blended behavioral analysis for identity theft detection in mobile social networks 1

引用

13th International Conference on Mobile Ad-hoc and Sensor Networks, MSN 2017

作者： Wang, Cheng Luo, Jing Yang, Bo Jiang, Changjun Key Laboratory of Embedded System and Service Computing Ministry of Education Department of Computer Science and Technology Tongji University Shanghai China

ISBN: (数字)9789811088902

ISBN: (纸本)9789811088896

User behavioral analysis is expected to act as a promising technique for identity theft detection in the Internet. The performance of this paradigm extremely depends on a good individual-level user behavioral model. Such a good model for a specific behavior is often hard to obtain due to the insufficiency of data for this behavior. The insufficiency of specific data is mainly led by the prevalent sparsity of users’ collectable behavioral footprints. This work aims to address whether it is feasible to effectively detect identify thefts by jointly using multiple unreliable behavioral models from sparse individual-level records. We focus on this issue in mobile social networks (MSNs) with multiple dimensions of collectable but sparse data of user behavior, i.e., making check-ins, posing tips and forming friendships. Based on these sparse data, we build user spatial distribution model, user post interest model and user social preference model, respectively. Here, as the arguments, we validate that there is indeed a complementary effect in multi-dimensional blended behavioral analysis for identity theft detection in MSNs. © Springer Nature Singapore Pte Ltd. 2018.

关键词： Social networking (online)

来源：评论

学校读者我要写书评

暂无评论

A Memetic Algorithm Based Feature Weighting for Metabolomics Data Classification

引用

Chinese Journal of Electronics 2014年第4期23卷 706-711页

作者： ZHOU Jiarui ZHU Zexuan JI Zhen College of Biomedical Engineering and Instrument Science Zhejiang University Shenzhen City Key Laboratory of Embedded System Design College of Computer Science and Software EngineeringShenzhen University

A metaheuristic chain based memetic algorithm namely MCMA is proposed for the classification of metabolomics data. MCMA regards both global evolution and local search as equivalent elemental metaheuristics,and searches with a chain of metaheuristics performed alternatively on the target problem. A hidden Markov model based scheduling mechanism is employed in MCMA for the selection of metaheuristics. By using MCMA for optimizing the linkage weight vector, a feature weighting algorithm for metabolomics data is formed to identify relevant metabolite features and reveal their exact relationships with the given physiological states. An extreme learning machine based classifier is utilized in predicting the physiological states according to the weighted metabolite features. Experimental results on real metabolomics data of clinical liver transplantation demonstrate that the proposed feature weighting and classification method obtains better performance than the other compared algorithms.

关键词： Metabolomics Feature weighting Memetic algorithm

来源：评论

学校读者我要写书评

暂无评论

引用

2007 International Conference on Parallel Processing Workshops, ICPPW 2007

作者： Zhang, Kailong Liang, Ke Zhou, Xingshe Wang, Kaibo Wu, Xiao Yang, Zhiyi School of Computer Northwestern Polytechnical University Shaanxi Key Embedded System Technology Laboratory

ISBN: (纸本)0769529348

Because of the resource constraints and high reliability requirement of embedded Distributed system (EDS), some new fault-tolerance means, which are different from the traditional hardware-redundancy ones, should be studied. In this article, a fault-tolerance method that based on similar resources and related technologies are proposed and discussed. First, several mathematical models of key elements, such as computing nodes, similar nodes and tasks, are constructed. Then, the similarity computation methods and evaluation criteria are evinced by two different views: tasks and resources. Supported by theories above, numerous methods, such as similar nodes auto-discovery (SNAD) and its optimization one (oSNAD), redundant tasks auto-deployment, and reconfiguration policies of fault tasks and nodes are highlighted respectively. Simulation results show that these approaches and schemes can improve the adaptive fault-tolerance abilities of complicated embedded distributed systems. © 2007 IEEE.

关键词： Fault tolerance

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：