检索结果-内蒙古大学图书馆

Super-resolution radial fluctuations (SRRF) algorithm with better real-time performance

INTERNATIONAL JOURNAL OF parallel EMERGENT AND distributed systems 2025年第3期40卷 201-213页

作者： Yang, Zhuhui Yang, Qiang Zhang, Gexiang Zhu, Ming Chengdu Univ Informat Technol Sch Automat Chengdu Peoples R China

Super-resolution radial fluctuations (SRRF) is a type of statistical analysis based on fluorescence fluctuation characteristics to achieve super-resolution imaging. At present, facing the problems of slow reconstruction speed and weak real-time image processing. For dealing with these issues, parallel computing characteristics of FPGA is employed in this paper to achieve the acceleration of the algorithm. Firstly, the reconstruction process of SRRF algorithm process is split into five sub-modules, which are designed depended on the pipeline design idea. Secondly, the top-level module is designed to accomplish the sub-modules connection for executing the SRRF algorithmic process, while the reconstruction is finished by using the high-density image dataset. Finally, the single-frame image processing time is recorded and analysed in FPGA with the mainstream configuration of PC. The experimental results show that, in terms of the single frame image processing time, SRRF in FPGA is improved by 77.95% compared to PC, with real-time image processing capability under 10 ms image sampling period. In addition, the resolution-scaled Pearson coefficient (RSP) and resolution-scaled error (RSE) values of the reconstruction results in FPGA are similar to those of PC.

关键词： Super-resolution radial fluctuations super-resolution imaging FPGA algorithm acceleration parallel computing real-time image processing

来源：评论

学校读者我要写书评

暂无评论

Advanced simulation-based predictive modelling for solar irradiance sensor farms

引用

JOURNAL OF SIMULATION 2025年第3期19卷 265-282页

作者： Risco-Martin, Jose L. Prado-Rujas, Ignacio-Iker Campoy, Javier Perez, Maria S. Olcoz, Katzalin Univ Complutense Madrid Dept Comp Architecture & Automat C Prof Jose Garcia Santesmases 9 Madrid 28040 Spain Univ Politecn Madrid Dept Comp Syst Architecture & Technol Madrid Spain

The need for accurate solar power forecasting is critical for grid stability as solar energy becomes more prevalent. This paper presents a new framework called Cloud-based Analysis and Integration for Data Efficiency (CAIDE) for real-time monitoring and forecasting of solar irradiance in sensor farms. CAIDE can handle multiple sensor farms, enhance predictive models in real-time, and is built on Model Based systems Engineering (MBSE) and Internet of Things (IoT) technologies. It can correct its forecasts, ensuring they stay current, and operates on various architectures, ensuring scalability. Tested on multiple sensor farms, CAIDE proved to be scalable and improved the initial accuracy of solar power production forecasts in real-time. This framework is significant for solar plant deployment and the advancement of renewable energy.

关键词： Complex systems discrete event system specification deep learning solar irradiance parallel and distributed simulation

来源：评论

学校读者我要写书评

暂无评论

An efficient distributed co-movement pattern detection framework for streaming trajectory

引用

KNOWLEDGE AND INFORMATION systems 2025年第6期67卷 5355-5375页

作者： Cheng, Tong Chao, Pingfu Zhang, Kenan Fang, Junhua Xu, Jiajie Soochow Univ Sch Comp Sci & Technol Suzhou Peoples R China

The ubiquitous of GPS-equipped devices has been generating massive trajectory data that record the movements of pedestrians, vehicles, and other moving objects. As one of the fundamental trajectory applications, the mining of co-movement patterns has been used extensively in location-based services (LBS), such as future motion forecasting and social recommendations. These applications require massive real-time data processing capabilities. However, the majority of existing research primarily focuses on historical data, while the rest minority targets the stream scenario, but only works on optimizing efficiency rather than scalability. In this paper, we focus on the distributed real-time co-movement pattern detection on trajectory stream. First, we propose a distributed streaming data processing framework based on Apache Flink. This framework consists of two stages: clustering and pattern mining. To accelerate clustering, we developed partitioning methods such as QGrid structure to achieve effective parallel processing of the spatial aspect of the clustering process. To efficiently perform pattern enumeration, we fully utilize the spatial information carried by cluster to narrow down the search space and perform verification effectively. Compared with the existing research on co-movement, we use spatial information to effectively integrate the two stages of data processing, reduce the search space of the pattern mining process, and improve the overall efficiency of the framework. Extensive experiments compared to existing methods have fully confirmed the efficiency of the proposed framework and its constituent technologies.

关键词： Trajectory distributed Clustering Quad-tree Pattern mining

来源：评论

学校读者我要写书评

暂无评论

Digital twin of dynamics for parallel kinematic machine with distributed force/position interaction

引用

JOURNAL OF MANUFACTURING systems 2025年 80卷 70-88页

作者： Zheng, Fangyan Han, Xinghui Hua, Lin Xu, Wenjun Wuhan Univ Technol Hubei Key Lab Adv Technol Automot Components Wuhan 430070 Peoples R China Wuhan Univ Technol Hubei Key Lab Broadband Wireless Commun & Sensor N Wuhan 430070 Peoples R China

Extensibility is significant for digital twin (DT) manufacturing systems. However, existing DT models mainly focus on a specific task in manufacturing. The main challenge lies in the specific physical model when addressing different tasks. In fact, the dynamics of machines are the physical basis for most applications, e.g., motion planning, production scheduling, process monitoring, machine maintenance, and so on. Therefore, the Digital Twin of dynamics (DTOD) for machines will be a foundation for a highly integrated and extensible DT system. However, due to the challenges in real-time dynamic modeling and the corresponding data interaction methods, the DTOD for parallel kinematic machines (PKM) has not been realized. Facing this challenge, this paper develops a DTOD for PKM with distributed force/position interaction. Firstly, a simplified rigid-flexible coupling dynamic model of PKM, considering link deformations, is established for realtime calculation. Then, a distributed position/force interaction method based on Kalman filter-based data fusion is proposed to realize high-performance data interaction between cyber and physical space. On this basis, a fivedimension digital twin model for DTOD of PKM is established. Further, the DTOD system with an architecture comprising dual central processors and multiple distributed edge executors/sensors is developed and validated by aircraft gear manufacturing, showing 80% prediction accuracy of dynamic error. Finally, to show the extensibility, integrated error correction for aircraft gear manufacturing is proposed as an extended application of the DTOD system. The gear error is reduced to 218 mu m (with error correction) from 503 mu m, representing a reduction of about 57 %, showing the high performance of the developed DTOD system and its high application potential.

关键词： Digital twin Physical model Data interaction parallel kinematic machine

来源：评论

学校读者我要写书评

暂无评论

Efficient distributed continual learning for steering experiments in real-time

引用

FUTURE GENERATION COMPUTER systems-THE INTERNATIONAL JOURNAL OF ESCIENCE 2025年 162卷

作者： Bouvier, Thomas Nicolae, Bogdan Costan, Alexandru Bicer, Tekin Foster, Ian Antoniu, Gabriel Univ Rennes CNRS Inria IRISA Rennes France Argonne Natl Lab Lemont IL USA

Deep learning has emerged as a powerful method for extracting valuable information from large volumes of data. However, when new training data arrives continuously ( i.e. , is not fully available from the beginning), incremental training suffers from catastrophic forgetting ( i.e. , new patterns are reinforced at the expense of previously acquired knowledge). Training from scratch each time new training data becomes available would result in extremely long training times and massive data accumulation. Rehearsal-based continual learning has shown promise for addressing the catastrophic forgetting challenge, but research to date has not addressed performance and scalability. To fill this gap, we propose an approach based on a distributed rehearsal buffer that efficiently complements data-parallel training on multiple GPUs to achieve high accuracy, short runtime, and scalability. It leverages a set of buffers (local to each GPU) and uses several asynchronous techniques for updating these local buffers in an embarrassingly parallel fashion, all while handling the communication overheads necessary to augment input minibatches using unbiased, global sampling. We further propose a generalization of rehearsal buffers to support both classification and generative learning tasks, as well as more advanced rehearsal strategies (notably Dark Experience Replay, leveraging knowledge distillation). We illustrate this approach with a real-life HPC streaming application from the domain of ptychographic image reconstruction. We run extensive experiments on up to 128 GPUs of the ThetaGPU supercomputer to compare our approach with baselines representative of training-from-scratch (the upper bound in terms of accuracy) and incremental training (the lower bound). Results show that rehearsal-based continual learning achieves a top-5 validation accuracy close to the upper bound, while simultaneously exhibiting a runtime close to the lower bound.

关键词： Continual learning Data-parallel training Experience replay distributed rehearsal buffers Asynchronous data management Scalability Streaming Generative AI

来源：评论

学校读者我要写书评

暂无评论

Co-Designing Transformer Architectures for distributed Inference With Low Communication

引用

IEEE TRANSACTIONS ON parallel AND distributed systems 2025年第4期36卷 717-730页

作者： Du, Jiangsu Wei, Yuanxin Ye, Shengyuan Jiang, Jiazhi Chen, Xu Huang, Dan Lu, Yutong Sun Yat Sen Univ Sch Comp Sci & Engn Guangzhou 510275 Peoples R China

Transformer models have shown significant success in a wide range of tasks. However, the massive resources required for its inference prevent deployment on a single device with relatively constrainted resources, thus leaving a high threshold of integrating their advancements. Observing scenarios such as smart home applications on edge devices and cloud deployment on commodity hardware, it is promising to distribute Transformer inference across multiple devices. Unfortunately, due to the tightly-coupled feature of Transformer model, existing model parallelism approaches necessitate frequent communication to resolve data dependencies, making them unacceptable for distributed inference, especially under relatively weak interconnection. In this paper, we propose DeTransformer, a communication-efficient distributed Transformer inference system. The key idea of DeTransformer involves the co-design of Transformer architecture to reduce the communication during distributed inference. In detail, DeTransformer is based on a novel block parallelism approach, which restructures the original Transformer layer with a single block to the decoupled layer with multiple sub-blocks. Thus, it can exploit model parallelism between sub-blocks. Next, DeTransformer contains an adaptive execution approach that strikes a trade-off among communication capability, computing power and memory budget over multiple devices. It incorporates a two-phase planning for execution, namely static planning and runtime planning. The static planning runs offline, containing a profiling procedure and a weight placement strategy before execution. The runtime planning dynamically determines the optimal parallel computing strategy from an expertly crafted search space based on real-time requests. Notably, this execution approach can adapt to heterogeneous devices by distributing workload based on devices' computing capabilities. We conduct experiments for both auto-regressive and auto-encoder tasks of Transforme

关键词： parallel processing Transformers Computational modeling Data models Planning distributed databases Runtime Hardware Adaptation models Throughput Model parallelism distributed transformer inference deep learning system

来源：评论

学校读者我要写书评

暂无评论

Task Allocation With Geography-Context-Capacity Awareness in distributed Burstable Billing Edge-Cloud systems

引用

IEEE TRANSACTIONS ON SERVICES COMPUTING 2025年第1期18卷 427-439页

作者： Shen, Shihao Gu, Chenfei Li, Yuanze Qiu, Chao Wang, Xiaofei Tan, Rui Zhang, Cheng Wang, Wenyu Tianjin Univ Coll Intelligence & Comp Tianjin 300350 Peoples R China Nanyang Technol Univ Sch Comp Sci & Engn Singapore 639798 Singapore Tianjin Univ Finance & Econ Inst Technol Tianjin 300222 Peoples R China PPIO Cloud Comp Shanghai Co Ltd Shanghai 200120 Peoples R China

The new real-time interactive services, such as virtual and augmented reality, demand significantly higher network bandwidth and quality, which the traditional centralized cloud struggles to meet. In addition, centralized optimization management becomes inefficient as the scale of the scene continues to expand. In response, edge cloud systems have emerged, but distributed geographic locations, burstable billing business models, and large numbers of servers in large-scale scenarios pose new challenges for resource management. In this article, we propose GeoCC, a novel strategy to save bandwidth overhead in burstable billing edge cloud systems. GeoCC addresses challenges through a dual approach. First, a geography-aware graph construction and partitioning algorithm is used to organize server resources, and a large number of servers are reasonably divided into multiple server pools for parallel processing. Second, it introduces an enhanced burstable billing optimization mechanism that considers contextual factors and adaptive bandwidth capacity. Experiments based on real data from an edge cloud operator demonstrate the effectiveness of GeoCC. Compared with the baseline, GeoCC can effectively reduce bandwidth peaks, decreasing bandwidth costs by an average of 28.30% and up to 81.83% at the 95th percentile billing.

关键词： Servers Bandwidth Cloud computing Costs Resource management Computer architecture real-time systems Optimization Biological system modeling Urban areas Edge-cloud architecture bandwidth optimization burstable billing server pools partition

来源：评论

学校读者我要写书评

暂无评论

Reinforcement Learning-Driven Adaptive Prefetch Aggressiveness Control for Enhanced Performance in parallel System Architectures

引用

IEEE TRANSACTIONS ON parallel AND distributed systems 2025年第5期36卷 977-993页

作者： Yang, Huijing Fang, Juan Hou, Yumin Su, Xing Xiong, Neal N. Beijing Univ Technol Coll Comp Sci Beijing 100124 Peoples R China Sul Ross State Univ Dept Comp Math & Phys Sci Alpine TX 79830 USA

In modern parallel system architectures, prefetchers are essential to mitigating the performance challenges posed by long memory access latencies. These architectures rely heavily on efficient memory access patterns to maximize system throughput and resource utilization. Prefetch aggressiveness is a central parameter in managing these access patterns;although increased prefetch aggressiveness can enhance performance for certain applications, it often risks causing cache pollution and bandwidth contention, leading to significant performance degradation in other workloads. While many existing prefetchers rely on static or simple built-in aggressiveness controllers, a more flexible, adaptive approach based on system-level feedback is essential to achieving optimal performance across parallel computing environments. In this paper, we introduce an Adaptive Prefetch Aggressiveness Control (APAC) framework that leverages Reinforcement Learning (RL) to dynamically manage prefetch aggressiveness in parallel system architectures. The APAC controller operates as an RL agent, which optimizes prefetch aggressiveness by dynamically responding to system feedback on prefetch accuracy, timeliness, and cache pollution. The agent receives a reward signal that reflects the impact of each adjustment on both performance and memory bandwidth, learning to adapt its control strategy based on workload characteristics. This data-driven adaptability makes APAC particularly well-suited for parallel architectures, where efficient resource management across cores is essential to scaling system performance. Our evaluation with the ChampSim simulator demonstrates that APAC effectively adapts to diverse workloads and system configurations, achieving performance gains of 6.73$\%$% in multi-core systems compared to traditional Feedback Directed Prefetching (FDP). By improving memory bandwidth utilization, reducing cache pollution, and minimizing inter-core interference, APAC significantly enhances pre

关键词： Prefetching systems architecture Bandwidth Pollution Accuracy System performance real-time systems Random access memory Resource management Adaptive systems Data prefetchers memory bandwidth prefetcher aggressiveness controller parallel system architectures reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

parallel Multi Objective Shortest Path Update Algorithm in Large Dynamic Networks

引用

IEEE TRANSACTIONS ON parallel AND distributed systems 2025年第5期36卷 932-944页

作者： Shovan, S. M. Khanda, Arindam Das, Sajal K. Missouri Univ Sci & Technol Dept Comp Sci Rolla MO 65409 USA

The multi objective shortest path (MOSP) problem, crucial in various practical domains, seeks paths that optimize multiple objectives. Due to its high computational complexity, numerous parallel heuristics have been developed for static networks. However, real-world networks are often dynamic where the network topology changes with time. Efficiently updating the shortest path in such networks is challenging, and existing algorithms for static graphs are inadequate for these dynamic conditions, necessitating novel approaches. Here, we first develop a parallel algorithm to efficiently update a single objective shortest path (SOSP) in fully dynamic networks, capable of accommodating both edge insertions and deletions. Building on this, we propose DynaMOSP, a parallel heuristic for Dynamic Multi Objective Shortest Path searches in large, fully dynamic networks. We provide a theoretical analysis of the conditions to achieve Pareto optimality. Furthermore, we devise a dedicated shared memory CPU implementation along with a version for heterogeneous computing environments. Empirical analysis on eight real-world graphs demonstrates that our method scales effectively. The shared memory CPU implementation achieves an average speedup of 12.74x and a maximum of 57.22x, while on an Nvidia GPU, it attains an average speedup of 69.19x, reaching up to 105.39x when compared to state-of-the-art techniques.

关键词： Heuristic algorithms Pareto optimization parallel algorithms Graphics processing units Wireless sensor networks Optimization Linear programming Aerodynamics Search problems Scalability Multi-objective shortest path dynamic graph GPU shared-memory SYCL programming model

来源：评论

学校读者我要写书评

暂无评论

distributed Joint Emitter Detection and Tracking With parallel Consensus on Likelihood and Prediction

引用

IET RADAR SONAR AND NAVIGATION 2025年第1期19卷

作者： Liang, Tao Zhang, Huaguo Wei, Ping Univ Elect Sci & Technol China Sch Informat & Commun Engn Chengdu Peoples R China

Passive localisation and tracking of a radio emitter is of significant interest for both civilian and defence applications. Among the existing methods, received signal strength indicator (RSSI)-based localisation is widely used due to its low cost and simplicity. However, most RSSI-based techniques make simplifying assumptions, such as relying on basic path-loss models and presuming the emitter has already been detected, overlooking complex environmental effects like shadowing caused by obstacles. These limitations hinder the practical application of RSSI-based methods. In this paper, we propose an advanced RSSI-based framework for joint detection and tracking (JDT) of a radio emitter, integrating a more accurate propagation model that accounts for both path-loss and shadowing effects. The emitter is modelled as a Bernoulli random finite set (RFS) characterised by its existence probability (EP) and spatial probability density function (SPDF), addressing the challenges of emitter detection uncertainty. A key innovation of this paper is the development of a fully distributed JDT algorithm, which overcomes the computational and communication challenges associated with centralised tracking systems. The proposed algorithm leverages parallel consensus on likelihood and prediction (PCLP), allowing for scalable and efficient operation across sensor networks. Simulation results validate the proposed method's performance in real-time emitter tracking.

关键词： distributed sensors radar detection RFS theory RSSI sensor fusion tracking

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：