您现在的位置:首页>外文期刊>Journal of supercomputing

期刊信息

  • 期刊名称:

    Journal of supercomputing

  • 中文名称: 超级计算杂志
  • 刊频: 0.687
  • ISSN: 0920-8542
  • 出版社: -
  • 简介:
  • 排序:
  • 显示:
  • 每页:
全选(0
<1/20>
3054条结果
  • 机译 使用预编码技术的基于认知无线电的MC-CDMA系统中的干扰消除
    摘要: In this article, the authors investigate error rate (ER) analysis of multi-carrier code-division multiple access (MC-CDMA) for cognitive radio network (CRN) using the pre-coding technique. CRN is a type of frequency-sensitive device in wireless communication which can intellect the idle spectrum availability and assign the spectrum dynamically for mobile communication. The spectrum is subdivided and its sub-band frequency of CRN is used for implementing multi-carrier (MC) communication and generating the spread of code frequency generation. In this work, pre-processing technique using singular value decomposition is considered for alleviation of the effects of interferences arising from secondary users to ensure a reliable link between the base station and the active users. The null space of channel matrix of active interference primary users has been exploited for the formulation of pre-processing matrix. A time-frequency domain signature sequence has been constructed to suppress the adverse effects of adjacent cell interference (ACI) and secondary multi-user interferences (S-MUI). Further, iterative decoder has been implemented for the achievement of better coding gain when the primary users' signals are contaminated by noise. Simulation ER results of CR MC-CDMA with pre-coding technique for Stanford University Interim and Long-term Evolution channels specification has been presented. It is discerned from ER results that coded CR MC-CDMA system established with pre-processing matrix offers robust performance through vindication of ACI and S-MUI effects while retaining a low complexity of primary mobile station for downlink transmission. Additionally, CRN-based MC-CDMA system has been analysed and found to provide an efficient and flexible solution to fulfil the demands of spectrum utilization.
  • 机译 NetGuru平台上的CSCL和TAoD平台上的CSPL用于网络实验班的认知学习成绩评估和分析
    摘要: How to find the best learning performance in the network experimental course is an important research topic. Therefore, collaborative learning and personalized learning have been applied in traditional classes for a similar significance regardless of how times change and advancements develop. We used the computer-supported collaborative learning (CSCL) pedagogy with NetGuru network experimental platform and the computer-supported personalized learning (CSPL) pedagogy with the teaching assistant on demand (TAoD) network experimental platform, respectively, in the "Network Engineering Lab (1)" course to assess the distinctions in impacts for the two methods. The objective students were four junior classes of ordinary students in the department of Computer Information and Network Engineering. As a rule, one should think about that CSCL pedagogy is better than CSPL pedagogy because of discussion and collaborative learning between students for learning performances. Nonetheless, the fact that the missing and late rates of students in technical and vocational college are a lot higher that makes adopting CSCL pedagogy difficult. Therefore, we designed and made some assistant materials of "graphical-based teaching powerpoint slides," "animations of packet transmission process for network protocol theories," and "Customized IP-oriented template-based assignments (simplified as CIPOTA or template-based assignment)" to be applied to compare and analyze the network practices of CSCL-based on NetGuru platform and CSPL-based on TAoD platform. We also design "satisfaction scale of learning performance" to analyze the students' satisfaction degrees of learning performance with the pedagogy (CSCL or CSPL), the assignment form (normal assignment or template-based assignment), and the learning satisfaction (CSCL or CSPL combines the normal assignment or template-based assignment). They were for the most part being utilized to demonstrate that TAoD platform-based CSPL pedagogy is truly appropriate for network experiment courses. And the CSPL pedagogy is not more terrible than the CSCL pedagogy and may even bring about optimizing learning results for the learning of network experiment course.
  • 机译 BOA:批处理编排算法,用于减轻异构GPU集群中分布式DL训练的拖累
    摘要: Training deep learning model is a time-consuming job since it usually uses a large amount of data. To reduce the training time, most practitioners train the models in a GPU cluster environment in a distributed way. The synchronous stochastic gradient descent, which is one of the widely used distributed training algorithms, has fast convergence rate with the use of multiple GPU workers; but its speed is tied to the slowest worker, i.e., straggler. In a heterogeneous environment, a static straggler, which has not been mainly focused before, has more impact on the performance than randomly occurring straggler. However, most existing studies for straggler mitigation usually consider a homogeneous environment, so their approaches are limited in practice. In this paper, we scrutinize the straggler problem under heterogeneous environment and define static and dynamic straggler from empirical results. Based on this, we propose a novel approach called batch orchestration algorithm (BOA) for straggler mitigation. It adaptively balances the amount of mini-batch data according to the speed of workers. Therefore, BOA can mitigate both static and dynamic straggler in a modern GPU cluster. BOA uses a Min-Max Integer programming to find the optimal mini-batch size, with the hardware-agnostic performance models. For verification, several experiments are conducted on a cluster having up to six GPUs with three types: GTX 1080, GTX 1060 and Quadro M2000. The results show BOA mitigates both types of stragglers and accelerates the training speed with synchronous SGD compared to other straggler mitigation method.
  • 机译 使用语义相似度方法对物联网环境进行垃圾邮件分类
    摘要: Unauthorized service or product advertising messages sent via electronic mails are called as spam e-mails. Detecting spam e-mail remains a challenging task. Existing countermeasures based on the statistical keyword, conceptual and IP address-based blacklists are not efficient due to difficulty in finding new attack patterns generated by the Internet of Things botnet devices. The other spam detection approaches rely on a hybrid of conceptual knowledge engineering with machine learning techniques. But, modern spammers evade the hybrid techniques through word polysemy and word ambiguity due to the context-sensitive nature of words. In this paper, the integration of Naive Bayesian classification with conceptual and semantic similarity technique is proposed to combat the ambiguity raised through polysemy in spam detection. To analyse the effectiveness of our approach, the experiments were conducted on benchmark data sets such as Spambase, PU1, Enron corpus, and Ling-spam. From the experimental results, it is evident that our proposed system achieves high accuracy of 98.89% than the existing approaches.
  • 机译 使用软件定义的网络范例的WMN的能量和QoS感知联合路由机制
    摘要: Providing energy conservation together with the quality of service (QoS) becomes more challenging in wireless mesh networks. A promising way to achieve this goal is to design a routing mechanism that takes both energy and QoS into account. On the other hand, software-defined networking (SDN) is an emerging network paradigm which separates the control plane from the data plane and facilitates a high level of programmability and manageability. In this paper, we propose FACOR, a novel SDN-based routing mechanism which aims at minimizing the energy consumption while providing a certain level of QoS for multimedia applications in wireless mesh networks. This technique is essentially suitable for battery-operated mesh nodes with solar panels. To obtain optimal paths, FACOR first estimates the cost of each network link using a fuzzy logic system. The fuzzy part takes parameters related to the quality of service and the energy in order to calculate the cost of the links. The routing problem is then formulated as an integer linear programming (ILP) model. Since ILP is NP-complete, we propose an ant colony optimization-based algorithm to solve the model. The proposed method is implemented in the well-known OpenDayLight controller and strictly follows a modular design for the sake of efficiency. Simulation results confirm the effectiveness of our mechanism which achieves a gain of about 17% and 10% in terms of network lifetime and PSNR, respectively, compared to other methods.
  • 机译 通过社交媒体数据和卫星图像进行土地利用分类
    摘要: Detailed urban land use classification plays a highly important role in the development and management of cities and in the identification of human activities. The complexity of the urban system makes its functional zoning a challenge, which makes such maps underutilized. A detailed land use classification encompasses both the natural land features and the classification of structures closely related to human activities. The use of satellite imagery to classify land use can effectively benefit the recognition of natural objects, but its performance demands significant improvement in the recognition of social functions due to the lack of information regarding human activities. To identify such activities in an urban area, we added Point of Interests (POI) data. This dataset contains both geographical tags and attributes that describe human activities. However, it has an uneven spatial distribution, with gaps in coverage being readily apparent. This paper proposes a land use classification framework using satellite imagery and data from social media. The proposed method employs a kernel density estimation to handle the spatial unevenness of POI data. The solution of mixed programming of MPI and OpenMP was adopted to parallel the algorithm. The results are compared to data compiled manually by means of human interpretation. Considering the example of Wuhan city, results show that the overall accuracy of land use type classification is 86.2%, and the Kappa coefficient is 0.860. It is demonstrated that using both POI and satellite images, a detailed land use map can be created automatically with satisfactory robustness.
  • 机译 通过统一结构和语义概念来提高异构多程序软件系统的模块化质量
    摘要: Program comprehension plays a significant role in the maintenance of software systems. There has recently been a significant increase in written large-scale applications with a collaboration of several programming languages. Due to the indirect collaboration between different components of a multilingual program, it is difficult to understand such program. Modularization is employed for extracting subsystems to help software system comprehension. The first step in the software modularization process is to extract a dependency graph from the source code. Taking into account all the programming languages used to implement a program, in the literature, there is no method to construct an integrated dependency graph from the source code aiming to support modularization of multilingual programs. To modularize such programs, we, first, create three dependency graphs named Call Dependency Graph (CDG), Semantic Dependency Graph (SDG) and Nominal similarity Dependency Graph (NDG) from the source code. The CDG, as a structural graph, is constructed for homogeneous programming languages and both SDG and NDG are built without taking into account the syntax of the programming languages used in the source code. Then, a genetic algorithm is presented to modularize multilingual programs from the constructed dependency graphs. The experimental results on Mozilla Firefox demonstrate that improvements in the simultaneous use of the SDG and NDG, and structural-based graph are 89%, 85%, 86%, and 59%, respectively, in terms of Precision, Recall, FM, and MoJoFM. The source codes and dataset related to this paper can be accessed at .
  • 机译 ElectricVIS:智慧城市供电数据可视化分析系统
    摘要: Smart grids provide a key driver for smart city development. The smart city power supply data visualization can realize the power characteristic information of various attributes and operating states in the online monitoring data of massive power equipments in a graphical and visual presentation, which provides a powerful guarantee for timely and effective monitoring and analysis of equipment operating status. However, with the rapid development of smart cities, the complexity of urban power data and the ever-increasing amount of data hinder the power managers' understanding and analysis of the power supply situation. Based on the smart city power supply data, a novel visual analysis system ElectricVis for urban power supply situation is proposed, which can interactively analyze large-scale urban power supply data. ElectricVis reduces the difficulty of understanding urban power supply situations by adopting novel visual graphic designs and time patterns that display power data in multiple scales. ElectricVis also provides different visual views and interaction methods for interrelated hierarchical data in urban power data, which is critical for detecting the cause of anomalous data. Finally, we evaluated our system through case studies and analysis by power experts.
  • 机译 混合调度平台:运行时预测可靠性感知调度平台,可提高HPC调度性能
    摘要: The performance of scheduling algorithms for HPC jobs highly depends on the accuracy of job runtime values. Prior research has established that neither user-provided runtimes nor system-generated runtime predictions are accurate. We propose a new scheduling platform that performs well in spite of runtime uncertainties. The key observation that we use for building our platform is the fact that two important classes of scheduling strategies (backfilling and plan based) differ in terms of sensitivity to runtime accuracy. We first confirm this observation by performing trace-based simulations to characterize the sensitivity of different scheduling strategies to job runtime accuracy. We then apply gradient boosting tree regression as a meta-learning approach to estimate the reliability of the system-generated job runtimes. The estimated prediction reliability of job runtimes is then used to choose a specific class of scheduling algorithm. Our hybrid scheduling platform uses a plan-based scheduling strategy for jobs with high expected runtime accuracy and backfills the remaining jobs on top of the planned jobs. While resource sharing is used to minimize fragmentation of resources, a specific ratio of CPU cores is reserved for backfilling of less predictable jobs to avoid starvation of these jobs. This ratio is adapted dynamically based on the resource requirement ratio of predictable jobs among recently submitted jobs. We perform extensive trace-driven simulations on real-world production traces to show that our hybrid scheduling platform outperforms both pure backfilling and pure plan-based scheduling algorithms.
  • 机译 基于并行计算的深度注意力模型用于命名实体识别
    摘要: Named entity recognition (NER) is an important task in natural language processing and has been widely studied. In recent years, end-to-end NER with bidirectional long short-term memory (BiLSTM) has received more and more attention. However, it remains a major challenge for BiLSTM to parallel computing, long-range dependencies and single feature space mapping. We propose a deep neural network model which is based on parallel computing self-attention mechanism to address these problems. We only use a small number of BiLSTMs to capture the time series of texts and then make use of self-attention mechanism that allows parallel computing to capture long-range dependencies. Experiments on two NER datasets show that our model is superior in quality and takes less training time. Our model achieves an F1 score of 92.63% on the SIGHAN bakeoff 2006 MSRA portion for Chinese NER, improving over the existing best results by over 1.4%. On the CoNLL2003 shared task portion for English NER, our model achieves an F1 score of 92.17%, which outperforms the previous state-of-the-art results by 0.91%.
  • 机译 基于虚拟六​​边形小区基础设施和多移动接收器的无线传感器网络节能数据分发算法
    摘要: In a wireless sensor network, reducing the energy consumption of sensor nodes and increasing energy conservation can be considered as two significant factors to prolong the network longevity. In this scheme, we proposed an energy-efferent data dissemination algorithm for the sensor nodes by making a virtual hexagonal cell-based infrastructure and multi-mobile sink to create a balance in energy consumption and to mitigate the energy-hole problem. The basic idea of the algorithm is to provide the ability for some of the sensing nodes that are elected by the virtual hexagonal backbone to inform other sensing nodes about the latest location of the nearest mobile sinks to decrease the overhead of updating the mobile sink location for the sensing nodes. Moreover, the sensor nodes can benefit considerably from these specified nodes as a relay node in the multi-hop routing process to send the data to the closest mobile sink based on the latest sink position information, which in turn leads to high energy conservation and low communication overhead between the sensing nodes and the mobile sinks in the network. According to the simulation results, the performance of the proposed scheme is better than the existing data transmission approaches in terms of total energy consumption, delay, and network longevity.
  • 机译 基于自主机器人自我意识的分布式任务分配方法
    摘要: Multi-robot task allocation problem is an important research field in multi-robot systems. The multi-robot task allocation problem combined with affective computing is an interesting frontier problem. This paper proposes a distributed affective robot pursuit task allocation algorithm based on self-awareness, which combines cognitive intelligence with emotional intelligence to establish self-awareness for affective robots. This paper defines the cognitive mechanism, environmental detection mechanism and self-decision mechanism of affective robots, so that the self-aware affective robots can make decisions autonomously by grasping the information of itself and the environment. The experimental results show better efficiency of this task allocation algorithm at each task scale, and the allocation efficiency is gradually improved as the number of robot increases.
  • 机译 弹性HDFS:互连的分布式架构,可增强大型云存储的可用性-可扩展性
    摘要: This paper presents an interconnected distributed architecture for storing data and metadata in large-scale cloud storage systems. The primary goal of the proposed architecture is to enhance the scalability of namespace directory in large-scale file systems. Structural shift from distinguished distributed model to interconnected distributed model and conducting effective coordination among file servers for namespace management are two key solutions considered in the context of proposed architecture. To this intent, a coordination protocol is designed for communication among file servers, and maintaining user transparency in the presence of different file system actions/reactions. The experimental results, obtained via emulations under different network conditions and cloud storage sizes, show up to 43.9% availability and 37.8% connection throughput improvements with negligible storage overhead compared to the latest released version of Hadoop distributed file system.
  • 机译 基于高性能计算的动态运行六足机器人
    摘要: Advanced high-performance computing (HPC) plays an important role in solving complex and large problems in robot simulation, parameter identification, and control. This article introduces a flexible hexapod robot with arc-shaped legs that has the ability to move fast and agilely based on HPC. Its mechanical structure design follows the principle of miniaturization and lightweight. While the robot uses drive devices with high power density, power supply, and drive, systems with high performance are designed for it to meet the requirements of high bursting capabilities and rapid movement capabilities. Meanwhile, this article proposes a gait generation and control method for the hexapod robot based on max-plus algebra. This method regards the movement of touching and leaving the ground during robot's walking as discrete events and uses a set of max-plus algebraic linear equations to describe the sequence of each discrete event in different locomotion gaits. By using this method, control laws for various gaits can be generated easily, switch of different gaits in real time can be achieved by switching the control matrix of these gaits, while the stability of the switching process and the synchronism of the locomotion of each leg before and after switching are also ensured. The virtual prototype simulation and physical prototype experiments were exerted to verify the effectiveness of the structure design and control method. Moreover, the simulation analysis and prototype experiment of Trotting diagonal gait and Pronking jumping gait were carried out in which the Pronking gait had a maximum traveling speed of 1.2 m/s.
  • 机译 GPU上的快速差分盒计数算法
    摘要: The differential box-counting (DBC) algorithm is the most widely used method for calculating the fractal dimension (FD) of grayscale images. FD analysis of grayscale images has important applications in fields such as shape classification, texture analysis and image segmentation. Nowadays these kinds of images can reach a very large size, especially in 3D, and there is a trend of increasing dataset size; this means that a current dataset of this type of image may need high computation times for computing the DBC. In this paper, we present an efficient implementation on graphics processing unit (GPU) of the original DBC algorithm in its optimized version for parallel processing. Our implementation with NVIDIA Compute Unified Device Architecture (CUDA) computes the DBC with high branch efficiency and very low GPU serialization. This fact allows us to obtain a very efficient GPU implementation of the algorithm. We tested our implementation on two different hardware/software platforms for a set of 2D and 3D grayscale images of increasing size. The results showed that this GPU implementation scaled very well and achieved a speedup of up to 52x with respect to a CPU single-thread implementation of the same algorithm. Against an OpenACC multi-thread implementation in CPU, our CUDA algorithm obtains a speedup of up to 6x.
  • 机译 使用统计分布式机器学习算法的生物特征识别的ECG数据优化
    摘要: Currently, security plays a crucial role in military, forensic and other industry applications. Traditional biometric authentication methods such as fingerprint, voice, face, iris and signature may not meet the demand for higher security. At present, the utility of biological signals in the area of security became popular. ECG signal is getting wide attention to use it as a tool for biometric recognition in authentication applications. ECG signals can provide more accurate biometrics for personal identity recognition. In machine learning, over-fitting is one of the major problems when working with a large data set of features so that an effective statistical technique is needed to control it. In this research, ECG signals are acquired from 20 individuals over 6 months in the MIT-BIH ECG-ID database. Altogether, a high-dimensional (N = 72) set of ECG features are extracted. These features are further fed to an algorithm, which reduces the feature space by classifying vital features and avoiding random, correlated and over-fitted features to increase the prediction accuracy. In this paper, a new intelligent statistical learning method, namely least absolute shrinkage and selection operator (LASSO), is proposed to select appropriate features for identification. The refined features thus obtained are trained with popular machine learning algorithms such as artificial neural networks, multi-class one-against-all support vector machine and K-nearest neighbour (K-NN). Finally, the performance of the proposed method with and without LASSO is compared using performance metrics. From the experimental results, it is observed that the proposed method of LASSO with K-NN classifier is effective with a recognition accuracy of 99.1379%.
  • 机译 使用内容和拓扑信息的鲁棒社区检测算法的优化工具
    摘要: With the recent prevalence of information networks, the topic of community detection has gained much interest among researchers. In real-world networks, node attribute (content information) is also available in addition to topology information. However, the collected topology information for networks is usually noisy when there are missing edges. Furthermore, the existing community detection methods generally focus on topology information and largely ignore the content information. This makes the task of community detection for incomplete networks very challenging. A new method is proposed that seeks to address this issue and help improve the performance of the existing community detection algorithms by considering both sources of information, i.e. topology and content. Empirical results demonstrate that our proposed method is robust and can detect more meaningful community structures within networks having incomplete information, than the conventional methods that consider only topology information.
  • 机译 基于自适应集体智能的差分进化算子
    摘要: In conventional differential evolutionary (DE) algorithm, mutation operator has significant influence on generating new vectors by mixing existing target vectors randomly selected from the current population. Recently, many mutation operators, which usually employ the best individual or some high-quality individuals randomly chosen, have been proposed to improve searching capability. However, such designs may easily suffer from premature convergence trapped by local optima. To make a trade-off between exploration and exploitation capability, this paper proposes a novel collective intelligence (CI)-based mutation operator, which is named as "current-to-sa-ci-best." In the presented mutation operator, the evolutionary information of m best target vectors is linearly combined to generate new mutant vectors. Besides, m is designed as an exponential-distributed random number which could be self-adapted based on successful records of m values alongside evolution. Moreover, this mutation operator could be applied to any DE algorithm without destroying existing search capability by adding a greedy selection operator. To verify its effectiveness, the proposed CI-based mutation strategy, which is named as SaCI, was embedded into some state-of-the-art DE variants on 28 CEC2013 benchmark functions. Numerical results have confirmed that the SaCI operator may be beneficial to DEs to some extent.
  • 机译 BP神经网络在人员岗位评价模型层次分析中的应用
    摘要: The matching of personnel and management of college administrators is related to the normal operation and development of colleges and universities. In view of the problem that person-post matching is too subjective in decision-making, the weight of each index in the evaluation index system of the manager's person-post matching is determined by using analytic hierarchy process (AHP) in this paper. Then, evaluation of person-post matching of managers by BP neural network and a person-post matching evaluation model of university managers based on BP neural network is constructed. The experimental results show that the error of person-post matching model is controlled within 5%, and a good evaluation result is obtained. At the same time, it shows that the model can adapt to the expert's brain thinking to process the sample data, fully absorb the expert's judgment experience, can scientifically and effectively evaluate the matching of managers and personnel, and provide a new idea for the objective decision-making of the person-post matching problem.
  • 机译 无线传感器网络环境中使用扩展DART和表消除(ET-DART)技术的安全分布式路由技术
    摘要: Distributed wireless sensor networks (WSNs) particularly deployed in urban applications like traffic surveillance are the main source of big data, and they produce a massive quantity of data. Multipath routing favors reliable data delivery in the case of sensitive data. However, the disadvantage is that many routes might increase the number of control packets. In this paper, we present an extended dynamic address routing technique (DART) with table elimination (ET) technique named ET-DART to reduce the complexity by handling a high degree of connectedness. The new idea of "connectedness" is used to distribute data from one node as it collected from another node. This method is used for defending capture attacks, and it modifies the security in every node without modifying the packet received ratios. A set of simulations takes place to verify the superiority of the presented ET-DART in terms of different performance metrics. Under the presence of 400 nodes, the presented ET-DART method successfully decrypts 154 packets with a maximum throughput of 82.15%. The experimental outcome verified the enhanced results of the presented method over the compared ones under several aspects.
  • 联系方式:010-58892860转803 (工作时间) 18141920177 (微信同号)
  • 客服邮箱:kefu@zhangqiaokeyan.com
  • 京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-1 六维联合信息科技(北京)有限公司©版权所有
  • 客服微信
  • 服务号