检索结果-内蒙古大学图书馆

32nd European Signal Processing Conference (EUSIPCO)

作者： Jung, Alexander Aalto Univ Dept Comp Sci Espoo Finland

ISBN: (纸本)9789464593617;9798331519773

A key challenge in federated learning applications is the statistical heterogeneity of local datasets. Clustered federated learning addresses this challenge by identifying clusters of local datasets that are approximately homogeneous. One recent approach to clustered federated learning is generalized total variation minimization (GTVMin). This approach requires a similarity graph which can be obtained by domain expertise or in a data-driven fashion via graph learning techniques. Under a widely applicable clustering assumption, we derive an upper bound the deviation between GTVMin solutions and their cluster-wise averages.

关键词： machine learning federated learning distributed algorithms convex optimization complex networks

来源：评论

学校读者我要写书评

暂无评论

Random Sleep Scheme-Based distributed Optimization Algorithm Over Unbalanced Time-Varying Networks

引用

IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS 2021年第8期51卷 5244-5253页

作者： Li, Huaqing Wang, Zheng Xia, Dawen Han, Qi Southwest Univ Chongqing Key Lab Nonlinear Circuits & Intelligen Coll Elect & Informat Engn Chongqing 400715 Peoples R China Guizhou Minzu Univ Coll Data Sci & Informat Engn Guiyang 550025 Peoples R China Guizhou Minzu Univ Coll Natl Culture & Cognit Sci Guiyang 550025 Peoples R China Chongqing Univ Sci & Technol Sch Intelligent Technol & Engn Chongqing 400715 Peoples R China

This article considers a category of constrained convex optimization problems over multiagent networks. The networked agents aim at collaboratively minimizing the sum of all locally known objective functions over a common convex set. Each agent possesses only its local convex function and its state is constrained to a privately known convex set. A novel distributed algorithm is proposed over time-varying unbalanced directed networks based on epigraph form of the original optimization problem and consensus theory. By incorporating the random sleep scheme, the proposed algorithm allows each agent to independently and randomly decide whether to calculate subgradient and take projection at each iteration, which alleviates the cost of subgradient observation. Besides, it neither resorts to doubly stochastic weight matrices (but only row-stochastic) nor the information of the graph sequence to execute. The convergence of the algorithm is explicitly analyzed under conditions that the sequence of time-varying directed graphs is uniformly jointly strongly connected and the subgradients of all local objective functions are bounded over a convex set. The optimization algorithm ensures zero-gap on the expected distance between the estimated value of each agent and the exact optimal solution. The two simulation cases are presented to demonstrate the practicability of the algorithm and correctness of the obtained theoretical results.

关键词： Optimization Convex functions distributed algorithms Linear programming Convergence Protocols Communication networks distributed convex optimization epigraph form random sleep time-varying unbalanced networks

来源：评论

学校读者我要写书评

暂无评论

Stochastic Gradient-Based distributed Bayesian Estimation in Cooperative Sensor Networks

引用

IEEE TRANSACTIONS ON SIGNAL PROCESSING 2021年 69卷 1713-1724页

作者： Cadena, Jose Ray, Priyadip Chen, Hao Soper, Braden Rajan, Deepak Yen, Anton Goldhahn, Ryan Lawrence Livermore Natl Lab Livermore CA 94550 USA Boise State Univ Dept Elect & Comp Engn Boise ID 83725 USA Voleon Grp Berkeley CA 94704 USA

distributed Bayesian inference provides a full quantification of uncertainty offering numerous advantages over point estimates that autonomous sensor networks are able to exploit. However, fully-decentralized Bayesian inference often requires large communication overheads and low network latency, resources that are not typically available in practical applications. In this paper, we propose a decentralized Bayesian inference approach based on stochastic gradient Langevin dynamics, which produces full posterior distributions at each of the nodes with significantly lower communication overhead. We provide analytical results on convergence of the proposed distributed algorithm to the centralized posterior, under typical network constraints. We also provide extensive simulation results to demonstrate the validity of the proposed approach.

关键词： Bayes methods Heuristic algorithms Convergence distributed databases Signal processing algorithms Inference algorithms Task analysis distributed algorithms Monte Carlo methods statistical learning autonomous systems network theory(graphs)

来源：评论

学校读者我要写书评

暂无评论

distributed Influence Maximization for Large-Scale Online Social Networks 38

Distributed Influence Maximization for Large-Scale Online So...

引用

38th IEEE International Conference on Data Engineering (ICDE)

作者： Tang, Jing Zhu, Yuqing Tang, Xueyan Han, Kai Hong Kong Univ Sci & Technol Data Sci & Analyt Thrust Hong Kong Peoples R China Nanyang Technol Univ Sch Comp Sci & Engn Singapore Singapore Univ Sci & Technol China Sch Comp Sci & Technol Hefei Peoples R China

ISBN: (数字)9781665408837

ISBN: (纸本)9781665408837

Thanks to billions of users in online social networks (OSNs), viral marketing becomes one of the most effective promotion channels for various new products or campaigns. Influence maximization is a classic problem in viral marketing, which has been extensively studied in the past two decades. Existing algorithms for influence maximization, however, mostly focus on single machine processing. To address the influence maximization problem on a massive scale, we design distributed algorithms via a cluster of machines, which can effectively speed up the computation while maintaining the state-of-the-art (1 - 1/e-epsilon)-approximation guarantee. Our distributed algorithms consist of two building blocks: (i) distributed reverse influence sampling, and (ii) element-distributed maximum coverage. We carry out extensive experiments on real datasets with millions of nodes and billions of edges to demonstrate the scalability of our distributed algorithms for both influence maximization and maximum coverage. In particular, our distributed algorithms accelerate the state-of-the-art IMM algorithm by 31x-56x times using a machine with 64 cores.

关键词： Influence maximization distributed algorithms Online social networks

来源：评论

学校读者我要写书评

暂无评论

Accelerated distributed Composite Nesterov Gradient Descent Algorithm 41

Accelerated Distributed Composite Nesterov Gradient Descent ...

引用

第41届中国控制会议

作者： Xinhui Yue Yuan-Hua Ni College of Artificial Intelligence Nankai University

ISBN: (数字)9789887581536

ISBN: (纸本)9781665482561

For the composite function composed of L-smooth general convex functions and non-smooth convex functions,a new accelerated distributed composite Nesterov gradient descent algorithm（Acc-DCNGD-NSC） is proposed in this *** algorithm can be applied to smooth convex optimization problems and composite convex optimization *** applying Nesterov acceleration techniques and a new gradient estimation scheme to the distributed proximal gradient algorithm,Acc-DCNGD-NSC is able to converge at a sub-linear rate of O（1/t）to global optimal solution.

关键词： distributed algorithms Nesterov accelerated algorithms proximal gradient algorithms gradient estimators

来源：评论

学校读者我要写书评

暂无评论

A Practical distributed ADMM Solver for Billion-Scale Generalized Assignment Problems 22

A Practical Distributed ADMM Solver for Billion-Scale Genera...

引用

31st ACM International Conference on Information and Knowledge Management (CIKM)

作者： Zhou, Jun Qi, Feng Hua, Zhigang Jian, Daohong Liu, Ziqi Wu, Hua Zhejiang Univ Coll Comp Sci Hangzhou Peoples R China Ant Grp Hangzhou Peoples R China Ant Grp Sunnyvale CA USA Ant Grp Beijing Peoples R China

ISBN: (纸本)9781450392365

Assigning items to owners is a common problem found in various real-world applications, for example, audience-channel matching in marketing campaigns, borrower-lender matching in loan management, and shopper-merchant matching in e-commerce. Given an objective and multiple constraints, an assignment problem can be formulated as a constrained optimization problem. Such assignment problems are usually NP-hard [21], so when the number of items or the number of owners is large, solving for exact solutions becomes challenging. In this paper, we are interested in solving constrained assignment problems with hundreds of millions of items. Thus, with just tens of owners, the number of decision variables is at billion-scale. This scale is usually seen in the internet industry, which makes decisions for large groups of users. We relax the possible integer constraint, and formulate a general optimization problem that covers commonly seen assignment problems. Its objective function is convex. Its constraints are either linear, or convex and separable by items. We study to solve our generalized assignment problems in the Bregman Alternating Direction Method of Multipliers (BADMM) framework where we exploit Bregman divergence to transform the Augmented Lagrangian into a separable form, and solve many subproblems in parallel. The entire solution can thus be implemented using a MapReduce-style distributed computation framework. We present experiment results on both synthetic and real-world datasets to verify its accuracy and scalability.

关键词： Assignment Problem ADMM Large-Scale Optimization distributed algorithms

来源：评论

学校读者我要写书评

暂无评论

The topology of local computing in networks

引用

Journal of Applied and Computational Topology 2024年第4期8卷 1069-1098页

作者： Fraigniaud, Pierre Paz, Ami Institut de Recherche en Informatique Fondamentale CNRS Université Paris Cité Laboratoire Interdisciplinaire des Sciences du Numérique CNRS Université Paris-Saclay

For more than three decades, distributed systems have been described and analyzed using topological tools, primarily using two techniques: protocol complexes and directed algebraic topology. In both cases, the considered computational model generally assumes communication via shared objects (typically a shared memory consisting of a collection of read-write registers) or message-passing enabling direct communication between any pair of processes. This paper aims to examine the use of protocol complexes in the study of network computing. In this case, processes are located at the network nodes and communicate by exchanging messages only along the network’s edges (i.e., not every pair of processes can directly communicate). There are several reasons why applying the topological approach to network computing can be challenging, and a prominent one is that node identifiers yield protocol complexes whose sizes grow exponentially with the size of the underlying network. However, many of the problems studied in this context are of local nature, and their definitions do not depend on the identifiers or the network size. We leverage this independence to meet the above challenge and present local protocol complexes, whose sizes do not depend on the network size. As an application of the “compacted” protocol complexes, we reformulate the celebrated lower bound of Ω(log^∗n) rounds for 3-coloring the n-node ring in the topological framework. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2024.

关键词： Combinatorial topology distributed algorithms distributed computing distributed graph algorithms Theory of computation

来源：评论

学校读者我要写书评

暂无评论

Partition Detection in Byzantine Networks 44

Partition Detection in Byzantine Networks

引用

44th IEEE International Conference on distributed Computing Systems (ICDCS)

作者： Bromberg, Yerom-David Decouchant, Jeremie Sourisseau, Manon Taiani, Francois Univ Rennes IRISA CNRS INRIA Rennes France Delft Univ Technol Delft Netherlands

ISBN: (纸本)9798350386066;9798350386059

Detecting and handling network partitions is a fundamental requirement of distributed systems. Although existing partition detection methods in arbitrary graphs tolerate unreliable networks, they either assume that all nodes are correct or that a limited number of nodes might crash. In particular, Byzantine behaviors are out of the scope of these algorithms despite Byzantine fault tolerance being an active research topic for important problems such as consensus. Moreover, Byzantine-tolerant protocols, such as broadcast or consensus, always rely on the assumption of connected networks. This paper addresses the problem of detecting partition in Byzantine networks (without connectivity assumption). We present a novel algorithm, which we call NECTAR, that safely detects partitioned and possibly partitionable networks and prove its correctness. NECTAR allows all correct nodes to detect whether a network could suffer from Byzantine nodes. We evaluate NECTAR's performance and compare it to two existing baselines using up to 100 nodes running real code, on various realistic topologies. Our results confirm that NECTAR maintains a 100% accuracy while the accuracy of the various existing baselines decreases by at least 40% as soon as one participant is Byzantine. Although NECTAR's network cost increases with the number of nodes and decreases with the network's diameter, it does not go above around 500KB in the worst cases.

关键词： Byzantine Fault Network Partition Digital Signatures distributed algorithms

来源：评论

学校读者我要写书评

暂无评论

distributed predefined-time constrained social cost minimization problem under the partial information setting

引用

SYSTEMS & CONTROL LETTERS 2024年 191卷

作者： Liu, Dongxin Nian, Xiaohong Li, Fan Cent South Univ Sch Automat Changsha 410075 Peoples R China

In this letter, the social cost minimization problem with coupled equality constraints is studied, where the agents endeavor to collaboratively minimize the sum of all local cost functions, and each local cost function depends on a global decision variable consisting of the decisions of all participants. In a setting where only partial information about one's own decisions and cost function is available, the first single-time scale predefined-time distributed algorithm based on the time-based-generator scheme is designed to tackle the problem. With the help of the Lyapunov stability theory, it is demonstrated that the algorithm can converge to the optimal solution in an arbitrarily specified time independent of the initial state and control parameters. Finally, the effectiveness of the proposed algorithm is verified by numerical simulations.

关键词： distributed algorithms Predefined-time convergence Social cost minimization Partial information setting Coupled constraints

来源：评论

学校读者我要写书评

暂无评论

A Linear Push-Pull Average Consensus Algorithm for Delay-Prone Networks

A Linear Push-Pull Average Consensus Algorithm for Delay-Pro...

引用

European Control Conference (ECC)

作者： Makridis, Evagoras Charalambous, Themistoklis Univ Cyprus Sch Engn Dept Elect & Comp Engn CY-1678 Nicosia Cyprus Aalto Univ Sch Elect Engn Dept Elect Engn & Automat Espoo 02150 Finland

ISBN: (纸本)9798331540920;9783907144107

In this paper, we address the average consensus problem of multi-agent systems for possibly unbalanced and delay-prone networks with directional information flow. We propose a linear distributed algorithm (referred to as RP-PAC) that handles asynchronous updates and time-varying heterogeneous information delays. Our proposed distributed algorithm utilizes a surplus-consensus mechanism and information regarding the number of incoming and outgoing links to guarantee state averaging, despite the imbalanced and delayed information flow in directional networks. The convergence of the RPPAC algorithm is examined using key properties of the backward product of time-varying matrices that correspond to different snapshots of the directional augmented network.

关键词： distributed algorithms push-pull consensus average consensus directed graphs time-varying heterogeneous delays

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：