检索结果-内蒙古大学图书馆

51st International conference on Very Large Data Bases, VLDB 2025

作者： Hajidehi, Milad Rezaei Sridhar, Sraavan Seltzer, Margo University of British Columbia Vancouver Canada

Graph partitioning plays a pivotal role in various distributed graph processing applications, including graph analytics, graph neural network training, and distributed graph databases. A "good" graph partitioner reduces workload execution time, worker imbalance, and network overhead. Graphs that require distributed settings are often too large to fit in the main memory of a single machine. This challenge renders traditional in-memory graph partitioners infeasible, leading to the emergence of streaming solutions. Streaming partitioners produce lower-quality partitions, because they work from partial information and must make premature decisions before they have a complete view of a vertex’s neighborhood. We introduce CUTTANA, a streaming graph partitioner that partitions massive graphs (Web/Twitter scale) with superior quality compared to existing streaming solutions. CUTTANA uses a novel buffering technique that prevents the premature assignment of vertices to partitions and a scalable coarsening and refinement technique that enables a complete graph view, improving the intermediate assignment made by a streaming partitioner. We implemented a parallel version for CUTTANA that offers nearly the same partitioning latency as existing streaming partitioners. Our experimental analysis shows that CUTTANA consistently yields better partitioning quality than state-of-the-art streaming vertex partitioners in terms of both edge-cut and communication volume metrics. We also evaluate the workload latencies that result from using CUTTANA and other partitioners in distributed graph analytics and databases. CUTTANA outperforms the other methods in most scenarios (algorithms, datasets). In analytics applications, CUTTANA improves runtime performance by up to 59% compared to various streaming partitioners (i.e., HDRF, Fennel, Ginger, HeiStream). In graph database tasks, CUTTANA results in higher query throughput by up to 23%, without hurting tail latency. © 2024, VLDB Endowment. All

关键词： Graph Databases

来源：评论

学校读者我要写书评

暂无评论

Relevance-Based Multi-User Data Compression for Fronthaul Rate Reduction in Cell-Free Massive MIMO Systems 14

Relevance-Based Multi-User Data Compression for Fronthaul Ra...

引用

14th International ITG conference on Systems, Communications and Coding, SCC 2025

作者： Danaee, Alireza Hassanpour, Shayan Wubben, Dirk Dekorsy, Armin University of Bremen Department of Communications Engineering Germany

ISBN: (纸本)9798331522896

In the Cell-Free massive Multiple-Input MultipleOutput (CF-mMIMO) systems, a large number of distributed users are simultaneously served by multiple Radio Access Points (RAPs). In the uplink, each RAP receives noisy observations from several users and must locally compress these signals before forwarding them to the corresponding Central processing Unit (CPU) via multiple fronthaul channels, each subject to a rate limitation. The challenge is to design the compressed signals at the RAPs such that the received signals at the CPU retain as much information as possible about the users (to be retrieved). To address this, we adopt compression techniques based on the Information Bottleneck (IB) principle to design local quantizers at the RAPs by ensuring an efficient balance between the informativity and compactness of the compressed signals. We discuss here two different compression schemes: one that processes the signals independently across fronthaul links and another that leverages the side information from previously retrieved signals at the CPU. By using side information, the latter generally provides a better tradeoff between the compression efficiency and performance, albeit with an increased complexity. Through numerical simulations, we demonstrate the effectiveness of both IB-based schemes compared to the conventional compression methods, showing their potential for improving fronthaul rate efficiency and overall system performance in typical digital transmission scenarios. © 2025 IEEE.

关键词： parallel processing systems

来源：评论

学校读者我要写书评

暂无评论

Exploiting Julia for parallel RBF-Based 3D Surface Reconstruction: A First Experience

Exploiting Julia for Parallel RBF-Based 3D Surface Reconstru...

引用

Euromicro conference on parallel, distributed and Network-Based processing

作者： Pasquale De Luca Ardelio Galletti Livia Marcellino Mario Pianese Science and Technologies Department Parthenope University of Naples Naples Italy

The 3D surface reconstruction is critical for various applications, demanding efficient computational approaches. Traditional Radial Basis Functions (RBFs) methods are limited by increasing data points, leading to slower execution times. Ad-dressing this, our study introduces an experimental parallelization effort using Julia, as well-known for high-performance scientific computing. We developed an initial sequential RBF algorithm in Julia, then expanded it to a parallel model, exploiting Multi-Threading to enhance execution speed while maintaining accuracy. This initial exploration into Julia's parallel computing capabilities shows marked performance gains in 3D surface reconstruction, offering promising directions for future research. Our findings affirm Julia's potential in computationally intensive tasks, with test results confirming the expected time efficiency improvements.

关键词：

来源：评论

学校读者我要写书评

暂无评论

FPGA-based Real-Time High-resolution image processing for rocket on-board processing 17

FPGA-based Real-Time High-resolution Image Processing for ro...

引用

17th IEEE International conference on Computer Research and Development, ICCRD 2025

作者： Zhou, Hui Wang, Xiaofeng Xie, Yujia Cao, Yongxiang Tang, Yu Huo, Yongchun Wen, Huan Zhao, Xiongbo Li, Chaoran National Key Laboratory of Science and Technology on Aerospace Intelligent Control Beijing Aerospace Automatic Control Institute Beijing China Beihang University Computer Science and Engineering Beijing China Beihang University National Superior College for Engineers Beijing China

ISBN: (纸本)9798331531881

The structure of the rocket-borne model is inherently complex, with processed images exhibiting high resolution and generating substantial amounts of data and calculations. Achieving robust real-Time computing on an embedded platform poses significant challenges due to strictly limited resources, power consumption constraints, and size limitations. Our review of rocket-borne applications reveals considerable variability in the design resources of different devices, indicating a need for expanded design approaches. Upon evaluating existing methods, we identified two primary drawbacks. First, certain operators within the high-resolution target detection model are difficult to parallelize, resulting in significant inference delays that hinder the ability to meet task requirements. Although existing methods have been extended, there remains significant potential for performance enhancement in core scheduling for poor acceleration. This paper proposes an optimized architecture for the target detection algorithm accelerator designed for high-resolution images, along with a novel highly parallel data pre-processing and post-processing module implemented on FPGA to address these issues. Compared to the ARM implementation, this architecture demonstrates an improved performance of 24.64x. Furthermore, to ensure flexible application across various rocket launch scenarios, we introduce an optimization structure for convolution, pooling, and fusion operators and a multi-core expansion optimization method. This approach yields a 1.29x improvement in computing unit utilization compared to state-of-The-Art multi-core scaling efforts. Finally, we assessed the accelerator architecture across multiple FPGA platforms, achieving a peak processing element utilization rate of 99.71% for a single core and layer. The overall computing efficiency, excluding the first layer, exceeded 90%. The peak computing power for the four cores reached 1638.4 GOPS, and the end-To-end computation time for

关键词： image enhancement

来源：评论

学校读者我要写书评

暂无评论

Human Pose Estimation Network Based On MultiHeatmap Fusion 5

Human Pose Estimation Network Based On MultiHeatmap Fusion

引用

5th Asia-Pacific conference on image processing, Electronics and Computers, IPEC 2024

作者： Zhao, Yangguang Lu, Jun College of Computer Science and Technology Heilongjiang University Key Laboratory of Database and Parallel Computingof Heilongjiang Province Haerbin China

ISBN: (纸本)9798350374407

Estimating human pose in complex multi-frame situations is a challenging task and has attracted intensive research by many researchers. Although 3D human pose estimation methods have achieved remarkable results in scenes based on single images, their performance often fails once these model transformations are applied to video sequences. Common problems with these models include inability to cope with motion blur, out-of-focus videos, and occlusion of human poses. In order to solve the above problems, this paper proposes a feature extraction and representation model MHMF, which is used in the feature extraction stage of the model. The initial features extracted by the backbone network HRNet-w32 are guided by the heat map to the attention layer, which improves the network's attention to important areas for predicting key points of human posture. At the same time, the integration of the aggregation heat map and the backbone network heat map improves the spatiotemporal consistency of the key points. In addition, to improve the accuracy of mesh pose estimation under occlusion, this paper proposes a transformer-based NewDSTformer model. By adjusting the structure of the Transformer encoder, increasing the encoder level and combining it with the dynamic progressive attention masking method. The model can adapt to different input situations, handle the positional relationship of local key points, and be able to perform accurate detection even under occlusion. It was evaluated on the 3DPW data set and improved the accuracy by $0.3 \%$, indicating that this paper effectively improved the performance of 3D human mesh reconstruction. © 2024 IEEE.

关键词： Mesh generation

来源：评论

学校读者我要写书评

暂无评论

Research and Suggestions for Evaluation Metrics of On-board Video Desensitization

Research and Suggestions for Evaluation Metrics of On-board ...

引用

Information Science, parallel and distributed Systems (ISPDS), International conference on

作者： Xiaolong Huang Te Hu Haojie Ji Yupeng Chen Wangdong Lu Peng He Guowei Sun Jingyan Wang State Key Laboratory of Intelligent Vehicle Safety Technology Key Laboratory of Modern Measurement & Control Technology Ministry of Education Beijing Information Science & Technology University Beijing China Key Laboratory of Modern Measurement & Control Technology Ministry of Education Beijing Information Science & Technology University Beijing China China Automotive Engineering Research Institute Co. Ltd Chongqing China Zhongtong Bus Holding Co. Ltd Liaocheng China

ISBN: (数字)9798350352719

ISBN: (纸本)9798350352726

Sensitive data leakage has become an urgent problem to be solved as more images based functionalities are being developed in vehicles. However, there is a scarcity of evaluation for on-board videos data desensitization. This research analyzes on-board video desensitization process including the image pre-processing stage, sensitive area localization stage and sensitive area desensitization stage. Considering that, this paper presents several evaluation methods and metrics of sensitive target detection performance, privacy-utility evaluation of video file metadata and image, so as to provide reference for the related research on the desensitization evaluation of on-board video.

关键词： Measurement Location awareness Information science Target recognition Current measurement distributed databases Object detection

来源：评论

学校读者我要写书评

暂无评论

Eco-FL: Adaptive Federated Learning with Efficient Edge Collaborative Pipeline Training 51

Eco-FL: Adaptive Federated Learning with Efficient Edge Coll...

引用

51st International conference on parallel processing (ICPP)

作者： Ye, Shengyuan Zeng, Liekang Wu, Qiong Luo, Ke Fang, Qingze Chen, Xu Sun Yat Sen Univ Sch Comp Sci & Engn Guangzhou Peoples R China

ISBN: (纸本)9781450397339

Federated Learning (FL) has been a promising paradigm in distributed machine learning that enables in-situ model training and global model aggregation. While it can well preserve private data for end users, to apply it efficiently on IoT devices yet suffer from their inherent variants: their available computing resources are typically constrained, heterogeneous, and changing dynamically. Existing works deploy FL on IoT devices by pruning a sparse model or adopting a tiny counterpart, which alleviates the workload but may have negative impacts on model accuracy. To address these issues, we propose Eco-FL, a novel Edge Collaborative pipeline based Federated Learning framework. On the client side, each IoT device collaborates with trusted available devices in proximity to perform pipeline training, enabling local training acceleration with efficient augmented resource orchestration. On the server side, Eco-FL adopts a novel grouping-based hierarchical architecture that combines synchronous intra-group aggregation and asynchronous inter-group aggregation, where a heterogeneity-aware dynamic grouping strategy that jointly considers response latency and data distribution is developed. To tackle the resource fluctuation during the runtime, Eco-FL further applies an adaptive scheduling policy to judiciously adjust workload allocation and client grouping at different levels. Extensive experimental results using both prototype and simulation show that, compared to state-of-the-art methods, Eco-FL can upgrade the training accuracy by up to 26.3%, reduce the local training time by up to 61.5%, and improve the local training throughput by up to 2.6x.

关键词： Federated learning edge intelligence pipeline parallelism parallel processing

来源：评论

学校读者我要写书评

暂无评论

Decoding the Granular Puzzle of Macromolecules: Efficient 3D Protein Structure Alignment in the Age of Big Data with Apache Spark

Decoding the Granular Puzzle of Macromolecules: Efficient 3D...

引用

2024 IEEE International conference on Big Data, BigData 2024

作者： Malysiak-Mrozek, Bozena Pawlowicz, Paulina Sunderam, Vaidy Hung, Che-Lun Kwiecien, Andrzej Mrozek, Dariusz Silesian University of Technology Department of Distributed Systems and Informatic Devices Gliwice Poland Silesian University of Technology Department of Applied Informatics Gliwice Poland Emory University Department of Computer Science AtlantaGA United States National Yang Ming Chiao Tung University Institute of Biomedical Informatics Taipei City Taiwan

ISBN: (纸本)9798350362480

Proteins are complex biological information granules that play a crucial role in various cellular processes within living organisms. processing 3D protein structures, which are the most informative from the biological point of view, is both intricate and time-consuming. In particular, performing 3D protein structure searches against large protein datasets involves identifying similarities and conducting structural alignments across numerous molecules (granules). This task demands advanced methods for matching identical and similar regions within protein structures and substantial computational resources to handle large collections of macromolecular data efficiently. In this paper, we present our parallel implementation of scalable 3D structural alignment on the Apache Spark big data platform. We describe a customized approach that leverages Spark data transformations within the data processing pipeline for the alignment process. Our experimental results demonstrate that this solution, tightly integrated with the Spark processing model, is both efficient and scalable, even with the increasing volume of protein structure data. © 2024 IEEE.

关键词： Information granules

来源：评论

学校读者我要写书评

暂无评论

Communication-Efficient Federated Learning With Compensated Overlap-FedAvg

引用

IEEE TRANSACTIONS ON parallel AND distributed SYSTEMS 2022年第1期33卷 192-205页

作者： Zhou, Yuhao Ye, Qing Lv, Jiancheng Sichuan Univ Coll Comp Sci Chengdu 610017 Peoples R China

While petabytes of data are generated each day by a number of independent computing devices, only a few of them can be finally collected and used for deep learning (DL) due to the apprehension of data security and privacy leakage, thus seriously retarding the extension of DL. In such a circumstance, federated learning (FL) was proposed to perform model training by multiple clients' combined data without the dataset sharing within the cluster. Nevertheless, federated learning with periodic model averaging (FedAvg) introduced massive communication overhead as the synchronized data in each iteration is about the same size as the model, and thereby leading to a low communication efficiency. Consequently, variant proposals focusing on the communication rounds reduction and data compression were proposed to decrease the communication overhead of FL. In this article, we propose Overlap-FedAvg, an innovative framework that loosed the chain-like constraint of federated learning and paralleled the model training phase with the model communication phase (i.e., uploading local models and downloading the global model), so that the latter phase could be totally covered by the former phase. Compared to vanilla FedAvg, Overlap-FedAvg was further developed with a hierarchical computing strategy, a data compensation mechanism, and a nesterov accelerated gradients (NAG) algorithm. In Particular, Overlap-FedAvg is orthogonal to many other compression methods so that they could be applied together to maximize the utilization of the cluster. Besides, the theoretical analysis is provided to prove the convergence of the proposed framework. Extensive experiments conducting on both image classification and natural language processing tasks with multiple models and datasets also demonstrate that the proposed framework substantially reduced the communication overhead and boosted the federated learning process.

关键词： Collaborative work Data models Training Servers Computational modeling Convergence Deep learning distributed computing federated learning overlap efficient communication

来源：评论

学校读者我要写书评

暂无评论

High-speed and Low-area Implementations of a Generic and parallel Integral image Architecture 20

High-speed and Low-area Implementations of a Generic and Par...

引用

20th IEEE Interregional NEWCAS conference (IEEE NEWCAS)

作者： Chehaitly, Mouhamad Lalevee, Andre Napoleon, Thibault Jridi, Maher L bISEN ISEN Yncrea Quest Brest France

ISBN: (纸本)9781665401050

Nowadays many applications of computer vision and image processing are based on Integral image (IIM) algorithm. Since the first use of IIM by Viola-Jones in the face detection algorithm, its computation is still a real challenge and receive a great attention in the research area especially in the field of real time applications. The IIM is greedy in hardware resources where the total number of additions increases with the image size and requires frequent memory accesses which result in the bottleneck effect when realized with CPU. To improve the processing efficiency, we propose a generic parallel/pipeline architectures with and without memory based on dual direction (by columns then by rows) or two levels of data flow oriented in IIM computing architecture. An interesting characteristic of the proposed architectures is the genericity of the calculation, according to the degree of parallelism (the maximum column size and the maximum row size that can be calculated in a clock cycle), the quantization of image values, on the one hand and on the other hand these architectures can be used in two functional modes. Based on the synthesis results of the implementation using Altera Quartus prime lite edition targeting an Intel/Altera Cyclone V - (FPGA), the proposed architecture achieves a high-throughput and low-area, when compared to the state-of-theart methods. More particularly, for 480x 640 image size, the proposed IIM architecture involves 765 logic registers, 643 slice LUT, and just 13,7 kbits and it operates at a maximum frequency of 182.28 MHz. These results show that this approach is one of the best candidates for portable applications that require high speed processing and low hardware resources usage.

关键词： Integral image parallel/Pipeline architecture FPGA

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：