检索结果-内蒙古大学图书馆

Enhancing Federated Learning Convergence With Dynamic Data Queue and Data-Entropy-Driven Participant Selection

IEEE INTERNET OF THINGS JOURNAL 2025年第6期12卷 6646-6658页

作者： Herath, Charuka Liu, Xiaolan Lambotharan, Sangarapillai Rahulamathavan, Yogachandran Loughborough Univ London Inst Digital Technol London E20 3BS England

Federated learning (FL) is a decentralized approach for collaborative model training on edge devices. This distributed method of model training offers advantages in privacy, security, regulatory compliance, and cost efficiency. Our emphasis in this research lies in addressing statistical complexity in FL, especially when the data stored locally across devices is not identically and independently distributed (non-IID). We have observed an accuracy reduction of up to approximately 10%-30%, particularly in skewed scenarios where each edge device trains with only 1 class of data. This reduction is attributed to weight divergence, quantified using the Euclidean distance between device-level class distributions and the population distribution, resulting in a bias term (delta(k)) . As a solution, we present a method to improve convergence in FL by creating a global subset of data on the server and dynamically distributing it across devices using a dynamic data queue-driven FL (DDFL). Next, we leverage Data Entropy metrics to observe the process during each training round and enable reasonable device selection for aggregation. Furthermore, we provide a convergence analysis of our proposed DDFL to justify their viability in practical FL scenarios, aiming for better device selection, a non-suboptimal global model, and faster convergence. We observe that our approach results in a substantial accuracy boost of approximately 5% for the MNIST dataset, around 18% for CIFAR-10, and 20% for CIFAR-100 with a 10% global subset of data, outperforming the state-of-the-art (SOTA) aggregation algorithms.

关键词： Data models Convergence Internet of Things distributed databases Accuracy Training Mathematical models Servers Adaptation models Data entropy fairness FL federated learning (FL) not identically and independently distributed (non-IID) not identically and independently distributed (non-IID)

来源：评论

学校读者我要写书评

暂无评论

Identifiability in Dynamic Acyclic Networks With Partial Excitation and Measurement

引用

IEEE TRANSACTIONS ON AUTOMATIC CONTROL 2025年第4期70卷 2305-2320页

作者： Cheng, Xiaodong Shi, Shengling Lestas, Ioannis Hof, Paul M. J. Van den Wageningen Univ & Res Dept Plant Sci Math & Stat Methods Biometris NL-6700 AA Wageningen Netherlands MIT Dept Chem Engn Cambridge MA 02139 USA Univ Cambridge Dept Engn Control Grp Cambridge CB2 1PZ England Eindhoven Univ Technol Dept Elect Engn Control Syst Grp NL- 5600 MB Eindhoven Netherlands

This article deals with dynamic networks in which the causality relations between the vertex signals are represented by linear time-invariant transfer functions (modules). Considering an acyclic network where only a subset of its vertices are measured and a subset of the vertices are excited, we explore conditions under which all the modules are identifiable on the basis of measurement data. Two sufficient conditions are presented, where the first condition concerns an identifiability analysis that needs to be performed for each vertex, while the second condition, based on the concept of tree/antitree covering, results from a graphical synthesis approach to allocate actuators and sensors in acyclic networks for achieving generic identifiability.

关键词： Power system dynamics Network topology Topology Actuators Sufficient conditions Sensors Q measurement Dynamic scheduling distributed databases Directed graphs Actuator allocation acyclic graphs dynamic networks identifiability sensor placement system identification

来源：评论

学校读者我要写书评

暂无评论

Task-Aware Data Selectivity in Pervasive Edge Computing Environments

引用

IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING 2025年第1期37卷 513-525页

作者： Koukosias, Athanasios Anagnostopoulos, Christos Kolomvatsos, Kostas Univ Thessaly Dept Informat & Telecommun Volos 38221 Greece Univ Glasgow Sch Comp Sci Glasgow City G12 8QQ Scotland

Context-aware data selectivity in Edge Computing (EC) requires nodes to efficiently manage the data collected from Internet of Things (IoT) devices, e.g., sensors, for supporting real-time and data-driven pervasive analytics. Data selectivity at the network edge copes with the challenge of deciding which data should be kept at the edge for future analytics tasks under limited computational and storage resources. Our challenge is to efficiently learn the access patterns of data-driven tasks (analytics) and predict which data are relevant, thus, being stored in nodes' local datasets. Task patterns directly indicate which data need to be accessed and processed to support end-users' applications. We introduce a task workload-aware mechanism which adopts one-class classification to learn and predict the relevant data requested by past tasks. The inherent uncertainty in learning task patterns, identifying inliers and eliminating outliers is handled by introducing a lightweight fuzzy inference estimator that dynamically adapts nodes' local data filters ensuring accurate data relevance prediction. We analytically describe our mechanism and comprehensively evaluate and compare against baselines and approaches found in the literature showcasing its applicability in pervasive EC.

关键词： Filters Peer-to-peer computing Sensors distributed databases Internet of Things Data models Uncertainty Real-time systems Predictive models Edge computing data filter data selectivity one-class support vector machines fuzzy inference

来源：评论

学校读者我要写书评

暂无评论

Privacy-Aware Data Acquisition Under Data Similarity in Regression Markets

引用

IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 2025年 PP卷 PP页

作者： Pandey, Shashi Raj Pinson, Pierre Popovski, Petar Aalborg Univ Dept Elect Syst Connect Sect DK-9220 Aalborg Denmark Imperial Coll London Dyson Sch Design Engn London SW7 2AZ England Tech Univ Denmark Dept Technol Management & Econ DK-2800 Lyngby Denmark Halfspace DK-1114 Copenhagen Denmark

Data markets facilitate decentralized data exchange for applications such as prediction, learning, or inference. The design of these markets is challenged by varying privacy preferences and data similarity among data owners. Related works have often overlooked how data similarity impacts pricing and data value through statistical information leakage. We demonstrate that data similarity and privacy preferences are integral to market design and propose a query-response protocol using local differential privacy (LDP) for a two-party data acquisition mechanism. In our regression data market model, we analyze strategic interactions between privacy-aware owners and the learner as a Stackelberg game over the asked price and privacy factor. Finally, we numerically evaluate how data similarity affects market participation and traded data value.

关键词： Data privacy distributed databases Data models Privacy Information leakage Training Cameras Pricing Games Data acquisition Collaborative learning information leakage mechanism design regression markets Stackelberg game

来源：评论

学校读者我要写书评

暂无评论

Rethinking RAN Architecture for Deep Fusion of AI and Communication in 6G

引用

IEEE WIRELESS COMMUNICATIONS 2025年第3期32卷 164-174页

作者： Li, Nan Wang, Yang Sun, Qi Li, Xiang Huang, Jinri Liu, Chunhui Huang, Yuhong Hu, Zhenping Han, Yantao I, Chih-lin Beijing Univ Posts & Telecommun Beijing Peoples R China China Mobile Res Inst Beijing Peoples R China China Mobile Res Inst Dept Wireless & Terminal Technol Beijing Peoples R China China Mobile Res Inst Wireless & Device Technol Res Off Beijing Peoples R China ZGC Inst Ubiquitous X Innovat & Applicat Beijing Peoples R China China Mobile Commun Grp Co Ltd Beijing Peoples R China China Mobile Commun Grp Co Ltd Sci & Innovat Planning Div Sci & Technol Innovat Dept Beijing Peoples R China

The deep integration of artificial intelligence (AI) with the radio access network (RAN) is envisioned to revolutionize mobile communications as we progress toward 6G. This article first explores the enhancement of the AI functional framework for intelligent RAN optimization and automation. It also investigates the role of RAN in enhancing connectivity, and in providing additional computing and data services which are essential for pervasive intelligence in the 6G era. Subsequently, we investigate the design principles of RAN architecture, and propose the potential RAN architecture that enables the deep integration of AI and communication. The potential impacts of standardization are also analyzed. To demonstrate the effectiveness of the proposed RAN architecture, we present a case study on the joint optimization of communication and computing for AI service performance assurance, supported by lab test results.

关键词： Artificial intelligence 6G mobile communication Computer architecture Optimization Radio access networks Data models Computational modeling distributed databases Quality of service Automation

来源：评论

学校读者我要写书评

暂无评论

MimoSketch: A Framework for Frequency-Based Mining Tasks on Multiple Nodes With Sketches

引用

IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING 2025年第3期37卷 1311-1324页

作者： Wu, Wenfei Xu, Yuchen Peking Univ Sch Comp Sci Beijing 100871 Peoples R China

In distributed data stream mining, we abstract a MIMO scenario where a stream of multiple items is mined by multiple nodes. We design a framework named MimoSketch for the MIMO-specific scenario, which improves the fundamental mining tasks of item frequency estimation, item size distribution estimation, heavy hitter detection, heavy change detection, and entropy estimation. MimoSketch consists of an algorithm design and a policy to schedule items to nodes. MimoSketch's algorithm applies random counting to preserve a mathematically proven unbiasedness property, which makes it friendly to the aggregate query on multiple nodes;its memory layout is dynamically adaptive to the runtime item size distribution, which maximizes the estimation accuracy by storing more items. MimoSketch's scheduling policy balances items among nodes, avoiding nodes being overloaded or underloaded, which improves the overall mining accuracy. Our prototype and evaluation show that our algorithm can improve the accuracy of five typical mining tasks by an order of magnitude compared with the state-of-the-art solutions, and the scheduling policy further promotes the performance in MIMO scenarios.

关键词： Data mining Accuracy Estimation Frequency estimation distributed databases Scheduling Machine learning algorithms Entropy Data analysis Schedules distributed data streams frequency-based mining tasks unbiased sketch scheduling policy

来源：评论

学校读者我要写书评

暂无评论

AIEA: An Asynchronous Influence-Based Evolutionary Algorithm for Expensive Many-Objective Optimization

引用

IEEE TRANSACTIONS ON CYBERNETICS 2025年第2期55卷 786-799页

作者： Wei, Feng-Feng Chen, Wei-Neng Zhang, Jun South China Univ Technol Sch Comp Sci & Engn Guangzhou 510006 Peoples R China South China Univ Technol State Key Lab Subtrop Buildingand Urban Sci Guangzhou 510006 Peoples R China Nankai Univ Coll Artificial Intelligence Tianjin 30071 Peoples R China Zhejiang Normal Univ Sch Comp Sci & Technol Jinhua 321004 Peoples R China Hanyang Univ Dept Elect & Elect Engn Ansan 15588 South Korea

In expensive multi/many-objective optimization problems (EMOPs), the expensive objectives are generally accessed through different simulation tools, leading to different evaluation latencies and unbearable computational time for serial optimization. One promising approach to improve efficiency is to perform simulation and build surrogates separately for each objective in parallel. However, how to improve the model accuracy and select promising candidates without global information are big challenges. To alleviate these problems, this article proposes an asynchronous influence-based SAEA (AIEA) based on the client-server model. Each client approximates an objective and the server takes charge for evolution. To adaptively select promising candidates, the influence degree is introduced in candidate selection, which is calculated in the objective space to judge which candidate has more beneficial influence for evolution. With the selected candidate, the most-uncertain-first strategy is devised in objective selection for asynchronous evaluations and model improvement. To handle incomplete objective values, the nearest neighbor inheritance is adopted for unevaluated objectives. Comprehensive experiments compared with five surrogate-assisted EAs demonstrate the global optimization and scalability of AIEA.

关键词： Optimization Computational modeling Predictive models Training data Data models Vectors Servers Program processors Evolutionary computation distributed databases Expensive optimization many-objective optimization problems parallel and distributed evolutionary computation surrogate-assisted evolutionary algorithm

来源：评论

学校读者我要写书评

暂无评论

Online Management for Edge-Cloud Collaborative Continuous Learning: A Two-Timescale Approach

引用

IEEE TRANSACTIONS ON MOBILE COMPUTING 2024年第12期23卷 14561-14574页

作者： Lin, Shaohui Zhang, Xiaoxi Li, Yupeng Joe-Wong, Carlee Duan, Jingpu Yu, Dongxiao Wu, Yu Chen, Xu Sun Yat Sen Univ Sch Comp Sci & Engn Guangzhou 510006 Peoples R China Hong Kong Baptist Univ Dept Interact Media Kowloon Tong Hong Kong Peoples R China Carnegie Mellon Univ Dept Elect & Comp Engn Pittsburgh PA 15213 USA Pengcheng Lab Dept Commun Shenzhen 518066 Peoples R China Shandong Univ Inst Intelligent Comp Sch Comp Sci & Technol Qingdao 266237 Peoples R China Dongguan Univ Technol Sch Cyberspace Secur Dongguan 523808 Peoples R China

Deep learning (DL) powered real-time applications usually need continuous training using data streams generated over time and across different geographical locations. Enabling data offloading among computation nodes through model training is promising to mitigate the problem that devices generating large datasets may have low computation capability. However, offloading can compromise model convergence and incur communication costs, which must be balanced with the long-term cost spent on computation and model synchronization. Therefore, this paper proposes EdgeC3, a novel framework that can optimize the frequency of model aggregation and dynamic offloading for continuously generated data streams, navigating the trade-off between long-term accuracy and cost. We first provide a new error bound to capture the impacts of data dynamics that are varying over time and heterogeneous across devices, as well as quantifying varied data heterogeneity between local models and the global one. Based on the bound, we design a two-timescale online optimization framework. We periodically learn the synchronization frequency to adapt with uncertain future offloading and network changes. In the finer timescale, we manage online offloading by extending Lyapunov optimization techniques to handle an unconventional setting, where our long-term global constraint can have abruptly changed aggregation frequencies that are decided in the longer timescale. Finally, we theoretically prove the convergence of EdgeC3 by integrating the coupled effects of our two-timescale decisions, and we demonstrate its advantage through extensive experiments performing distributed DL training for different domains.

关键词： Data models Training Computational modeling Accuracy Costs distributed databases Federated learning Collaborative federated learning continuous learning edge-cloud collaboration two-timescale

来源：评论

学校读者我要写书评

暂无评论

Order Optimal Bounds for One-Shot Federated Learning Over Non-Convex Loss Functions

引用

IEEE TRANSACTIONS ON INFORMATION THEORY 2024年第4期70卷 2807-2830页

作者： Sharifnassab, Arsalan Salehkaleybar, Saber Golestani, S. Jamaloddin Univ Alberta Comp Sci Dept Reinforcement Learning & Artificial Intelligence R Edmonton AB T6G 2E8 Canada Leiden Univ Leiden Inst Adv Comp Sci LIACS NL-2311 EZ Leiden Netherlands Sharif Univ Technol Dept Elect Engn Tehran 113659466 Iran

We consider the problem of federated learning in a one-shot setting in which there are m machines, each observing n sample functions from an unknown distribution on non-convex loss functions. Let F:[-1, 1](d )-> R be the expected loss function with respect to this unknown distribution. The goal is to find an estimate of the minimizer of F . Based on its observations, each machine generates a signal of bounded length B and sends it to a server. The server collects signals of all machines and outputs an estimate of the minimizer of F . We show that the expected loss of any algorithm is lower bounded by max(1 root n-(mB)(1/d)), 1 root mn) , up to a logarithmic factor. We then prove that this lower bound is order optimal in m and n by presenting a distributed learning algorithm, called Multi-Resolution Estimator for Non-Convex loss function (MRE-NC), whose expected loss matches the lower bound for large mn up to polylogarithmic factors.

关键词： Servers Federated learning Manganese Computational modeling distributed databases Machinery Training distributed learning communication efficiency non-convex optimization

来源：评论

学校读者我要写书评

暂无评论

Accelerating distributed Repartition Joins on Skewed Datasets via Patch-Based Shuffling

引用

IEEE ACCESS 2025年 13卷 41068-41096页

作者： Kassela, Evdokia Konstantinou, Ioannis Koziris, Nectarios Natl & Tech Univ Athens Sch Elect & Comp Engn Athens 11527 Greece Univ Thessaly Dept Informat & Telecommun Lamia 35100 Greece

In distributed workloads involving joins and aggregations, skewed attribute values often cause load balancing issues, leading to stragglers and increased execution times. Existing solutions often rely on cost-based models, require extensive parameter tuning, or necessitate modifications to distributed execution engines, limiting their usability and generality. To address this challenge, we present a novel patch-based repartitioning algorithm that eliminates load imbalances while minimizing network overhead. Our approach extends the subset-replicate technique by leveraging data distribution and location statistics to optimize data locality and reduce unnecessary data movement. Unlike traditional hash-based methods, our technique is skew-insensitive, requires no parameter tuning, and integrates seamlessly into existing distributed execution engines as a drop-in replacement for the shuffle mechanism. The proposed method operates in three distinct stages: (1) statistics collection and load threshold computation, (2) patch-based subgroup assignment to ensure optimal load balancing with minimal replication, and (3) informed data shuffling and join execution. This structured process ensures even workload distribution across workers while reducing network I/O. Theoretical analysis proves the scalability, skew robustness, and load-balancing guarantees of our approach, establishing bounds on maximum worker load and network data movement. Experimental evaluations demonstrate that our method achieves perfectly balanced workloads and reduces execution time by up to 81% compared to conventional hash-based joins under moderate to high skew, while introducing negligible overhead at low skew levels. These improvements are attributed to reduced data movement, optimized worker utilization, and the algorithm's robust theoretical foundation. Our research provides a versatile and practical solution for skewed data processing, significantly advancing the efficiency of distributed data man

关键词： Engines Partitioning algorithms Load management distributed databases Costs Telecommunication traffic Optimization Logic Load modeling Histograms Load balancing repartition join shuffling subset-replicate

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：