检索结果-内蒙古大学图书馆

Annual IEEE Northeast workshop on Circuits and systems (NEWCAS)

作者： Liyuan Guo Seyed Mohammad Ali Zeinolabedin Franz Marcus Schüffny Annika Weiße Stefan Scholze Richard George Johannes Partzsch Christian Mayr Chair of Highly-Parallel VLSI-Systems and Neuro-Microelectronics Circuits Technische Universität Dresden Germany

The real-time multi-channel intracranial recording of neural signals is required in both neuroscientific research and clinical practice. Due to the limited power budget and the increasing number of recording channels, a compression engine is highly recommended. This paper proposes a 16-channel real-time adaptive compression engine (ACE) exploiting neural signal properties. It is able to switch between lossless and near-lossless compression modes. In near-lossless mode, it compresses the spike region lossless and discards the rest. The achieved space-saving ratio (SSR) is on average about 62.5% and 91% for lossless and near-lossless modes, respectively. It can save about 78.5% of the power consumption (near-lossless compression) compared to the transmission of raw neural signals. The 16-channel ACE is implemented in 22nm FDSOI technology and consumes 230.0 μW dynamic- and 55.49 μW leakage-power at 5 MHz.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Multi-Node Inference Architectures for Low-Latency LLM Serving

Multi-Node Inference Architectures for Low-Latency LLM Servi...

引用

Advanced Computing Technologies (ICoACT), International Conference on

作者： Naresh Kumar Gundla Sri Harsha Atthuluri

ISBN: (数字)9798331509859

ISBN: (纸本)9798331509866

Over the past few years, large language models have evolved to enable a wide range of applications-from natural language understanding to real-time conversational agents. However, the deployment of LLMs into production presents many significant challenges, especially with regard to low-latency responses that enable real-time interactions. This work investigates multi-node inference architectures for optimized deployment using open-source frameworks with scalability, flexibility, and cost-effectiveness. We investigate various methods, such as microbatching, tensor and pipeline parallelism, and sophisticated load balancing, that effectively distribute inference workloads across multiple nodes. We conduct extensive evaluations using popular open-source tools such as Kubernetes, Ray, and Envoy to benchmark the performance of these architectures in terms of latency, throughput, and resource utilization under diverse workloads. We also analyze model replication versus model partitioning trade-offs, giving insights into the most appropriate configuration for various deployment scenarios. As our results show, a well-orchestrated multi-node setup can be used to greatly reduce inference latency while preserving high throughputs, enabling the deployment of sophisticated LLMs in latencysensitive applications. This paper gives insights with a detailed analysis of multi-node inference strategies and integration into open-source ecosystems, therefore it will be a great guide for practitioners seeking to develop deployments of LLMs at scale. In summary, this work underlines how distributed architectures can overcome some of the inherent limitations imposed by singlenode deployments and are crucial for achieving more efficient and responsive AI-driven services.

关键词： Tensors Large language models Scalability Pipelines Computer architecture parallel processing Throughput real-time systems Low latency communication Load modeling

来源：评论

学校读者我要写书评

暂无评论

On the exponential leader-tracking control for high-order multi-agent systems via distributed PI strategy in the presence of heterogeneous time-varying delays 16th

On the exponential leader-tracking control for high-order mu...

引用

16th International-Federation-of-Automatic-Control (IFAC) workshop on time Delay systems (TDS)

作者： Caiazzo, Bianca Lui, Dario Giuseppe Petrillo, Alberto Santini, Stefania Univ Naples Federico II Dept Informat Technol & Elect Engn DIETI I-80125 Naples Italy

This paper deals with the exponential leader-tracking consensus control problem of high-order Multi-Agent systems (MASs) sharing information via a non-ideal communication network. To emulate a more realistic environment, a specific time-varying delay has been associated with each communication link within the network, whose value, at each time instant, depends on the real conditions of the communication channel. To solve this problem, a fully-distributed delayed Proportional-Integral (PI) control protocol able to guarantee the exponential stability of the entire delayed closed-loop MAS is proposed. The stability of this latter is analytically proved by exploiting Lyapunov-Krasovskii theory combined with Halanay Inequality, thus obtaining exponential stability conditions expressed as a feasible Linear Matrix Inequality (LMI) problem. Exemplary numerical simulations corroborate the effectiveness of the theoretical derivation. Copyright (C) 2021 The Authors.

关键词： Multi-Agent systems distributed PI control strategy Heterogeneous time-varying delays Lyapunov-Krasovskii theory Halanay Inequality

来源：评论

学校读者我要写书评

暂无评论

Differential evolution with stage stratification method and dual balanced mutation strategy for real-parameter numerical optimization

引用

EXPERT systems WITH APPLICATIONS 2024年第PartB期238卷

作者： Sun, Yu Yang, Guanxiong Guangxi Univ Sch Comp & Elect Informat Nanning 530004 Peoples R China Guangxi Univ Educ Dept Guangxi Zhuang Autonomous Reg Key Lab Parallel Distributed & Intelligent Comp Nanning Peoples R China Guangxi Univ Guangxi Key Lab Multimedia Commun & Network Techno Nanning 530004 Peoples R China

Traditional differential evolution (DE) algorithms have functional limitations in effectively addressing in-creasingly intricate numerical optimization problems. The key to responding to this challenge is to strike a suitable balance between exploration and exploitation. Exploration is used to find the global optimal solution, and exploitation is used to improve the accuracy of the global optimal solution. Therefore, this study introduces a novel differential evolution algorithm with a stage stratification method and a dual balanced mutation strategy framework, named SbmDE. To enhance the balance between convergence and diversity, the population is stratified into exploration and exploitation layers based on the size of the fitness value and evolutionary period. In the exploitation layer, a hybridization mutation is applied as the first mutation. A novel dual improved mutation operation is proposed and applied to the exploration layer. For the first mutation, the hybridization mutation is used in the same manner as for the exploitation layer. For the secondary mutation, two improved mutation strategies are proposed to alleviate premature convergence at the early stage of evolution and to enhance the local neighborhood search at a later stage, named DE/ranking-to-rand/1 and DE/best-current-dev/1. Experiments were conducted on the CEC2017 benchmark suite, which contains 29 single-objective real-parameter numerical optimization problems, to evaluate the performance of the proposed algorithm. Compared with 11 state-of-the-art algorithms, the results demonstrate the superiority of the proposed algorithm, which does not increase the time complexity. Additionally, based on the four engineering design problems, the proposed algorithm is fully competent in solving practical constrained optimization problems.

关键词： Differential evolution Stage stratification Dual mutation Balanced mutation real-parameter numerical optimization

来源：评论

学校读者我要写书评

暂无评论

OptimizeLog: Log Anomaly Detection and Localization based on Optimized Log Parsing in distributed systems

OptimizeLog: Log Anomaly Detection and Localization based on...

引用

International Conference on Computer and Communications (ICCC)

作者： Yanjie Sun Yali Gao Xiaoyong Li Key Laboratory of Trustworthy Distributed Computing and Service (BUPT) Ministry of Education Beijing University of Posts and Telecommunications Beijing China

Modern distributed systems generate interleaved logs when performing parallel operations, and these logs become an important basis for anomaly detection and localization. To achieve more robust and accurate log anomaly detection and localization, we propose a comprehensive model, namely OptimizeLog. We built a dictionary- and length-based log parser that works stably in most log systems without parameter tuning. We significantly improve anomaly detection accuracy by introducing ALBERT-based semantic embeddings, count embeddings, and time embeddings, combined with an attentionbased Bi-GRU model. By constructing component instance forest and tree-based depth-first traversal algorithm, abnormal location in distributed systems is realized. Experimental results show that our log parsing method is 4% more accurate than other log parsers. Compared with other advanced methods such as DeepLog, LogAnomaly, and LogRobust, OptimizeLog improves the accuracy of anomaly detection by 5%, while enabling instancelevel anomaly localization on real datasets.

关键词：

来源：评论

学校读者我要写书评

暂无评论

distributed Model Predictive Control for Multiple Trains with Coupled Constraints

Distributed Model Predictive Control for Multiple Trains wit...

引用

第42届中国控制会议

作者： Zixuan Zhang Haifeng Song Weiqi Bai Hongwei Wang Hairong Dong State Key Laboratory of Rail Traffic Control and Safety Beijing Jiaotong University School of Electronic Information Engineering Beihang University School of Automation Science and Electrical Engineering Beihang University

This paper presents a distributed model predictive control of multiple trains,aiming at overcoming coupled constraints and achieving cooperative operation while reducing the computational *** to the coupled constraints associated with maintaining safety distance between adjacent trains,trains are divided into sub-systems and distributed optimal control problem is then *** solve the problem,an alternating direction method of multipliers(ADMM) is employed with a fully distributed scheme,decomposing the original optimal control problem into smaller sub-problems that can be computed in parallel and meet the real-time control *** proposed method effectively guarantees the cooperative operation of trains under local and coupled safety *** simulations are provided to demonstrate the feasibility and effectiveness of the proposed train control model and methods.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Visualizing Kubernetes distributed systems: An Exploratory Study

Visualizing Kubernetes Distributed Systems: An Exploratory S...

引用

IEEE International workshop on Visualizing Software for Understanding and Analysis (VISSOFT)

作者： Dennis Giovani Balreira Thiago da Silva Araújo Fabio Petrillo Institute of Informatics UFRGS Porto Alegre Brazil Departement de genie logiciel et des TI Ecole de Technologie Superieure Montreal Canada

distributed applications running on virtualization- based systems and cloud computing have become popular solutions, allowing developers to focus on application logic rather than dealing with the complexities of distributed systems. However, these applications often become increasingly complex, presenting multiple management challenges. To address this issue, software visualization approaches offer valuable solutions by pro-viding real-time insights into resources and their functionalities, offering a comprehensive overview. This study aims to analyze and evaluate existing software visualization tools for distributed applications on the Kubernetes platform. The objective is to comprehensively examine these tools' features, capabilities, and limitations to understand their effectiveness in visualizing complex distributed systems. Our findings provide valuable insights into the strengths and weaknesses of the available visualization tools, enabling researchers and practitioners to make informed decisions and advancements in software visualization for distributed applications on the Kubernetesplatform. Our research identified eight Kubernetes visualization tools, which were examined and compared based on relevant char-acteristics related to distributed applications and software vi-sualization standards. However, it is worth noting that despite the excellent work done by the community in establishing these first proposals, these tools currently only support, on average, a visualization of 9 % of the total resource types available, as mentioned in the official documentation. Therefore, we propose guidelines followed by a synthesized visualization that can guide further research and development in this area. Our study will assist users in selecting the most suitable Kubernetes visualization tool and encourage researchers and the community to explore new approaches in Kubernetes visualization.

关键词：

来源：评论

学校读者我要写书评

暂无评论

distributed On-Demand Deployment for Transparent Access to 5G Edge Computing Services

Distributed On-Demand Deployment for Transparent Access to 5...

引用

IEEE International Symposium on parallel and distributed Processing workshops and Phd Forum (IPDPSW)

作者： Josef Hammer Hermann Hellwagner Institute of Information Technology Alpen-Adria-Universität Klagenfurt Austria

Multi-access Edge Computing (MEC) is a central piece of 5G telecommunication systems and is essential to satisfy the challenging low-latency demands of future applications. MEC provides a cloud computing platform at the edge of the radio access network. Our previous publications argue that edge computing should be transparent to clients, leveraging Software-Defined Networking (SDN). While we introduced a solution to implement such a transparent approach, one question remained: How to handle user requests to a service that is not yet running in a nearby edge cluster? One advantage of the transparent edge is that one could process the initial request in the cloud. However, this paper argues that on-demand deployment might be fast enough for many services, even for the first request. We present an SDN controller that automatically deploys an application container in a nearby edge cluster if no instance is running yet. In the meantime, the user’s request is forwarded to another (nearby) edge cluster or kept waiting to be forwarded immediately to the newly instantiated instance. Our performance evaluations on a real edge/fog testbed show that the waiting time for the initial request – e.g., for annginx-based service – can be as low as 0.5 seconds – satisfactory for many applications.

关键词：

来源：评论

学校读者我要写书评

暂无评论

distributed Optimization of District Heating Networks Using Optimality Condition Decomposition

引用

ENERGIES 2022年第18期15卷 6605页

作者： Maurer, Jona Illerhaus, Jochen Soneira, Pol Jane Hohmann, Soeren Karlsruhe Inst Technol KIT Inst Control Syst IRS Kaiserstr 12 D-76131 Karlsruhe Germany

The optimal operation of District Heating Networks (DHNs) is a challenging task. Current or future optimal dispatch energy management systems attempt to optimize objectives, such as monetary cost minimization, emission reduction, or social welfare maximization. Typically, this requires highly nonlinear models and has a substantial computational cost, especially for large DHNs. Consequently, it is difficult to solve the resulting nonlinear programming problem in real time. In particular, as typical applications allow for no more than several minutes of computation time. However, a distributed optimization approach may provide real time performance. Thereby, the solution of the central optimization problem is obtained by solving a set of small-scale, coupled optimization problems in parallel. At runtime, information is exchanged between the small-scale problems during the iterative solution procedure. A well-known approach of this class of distributed optimization algorithms is Optimality Condition Decomposition (OCD). Important advantages of this approach are the low amount of information exchange needed between the small-scale problems and that it does not require the tuning of parameters, which can be challenging. However, the DHNs model equation structure brings along many difficulties that hamper the application of the OCD approach. Simulation results demonstrate the applicability range of the presented method.

关键词： district heating networks optimal dispatch nonlinear optimization distributed optimization optimality condition decomposition electric power networks

来源：评论

学校读者我要写书评

暂无评论

Cooperative Object Transportation Using Autonomous Networked Cobots: A distributed Approach

Cooperative Object Transportation Using Autonomous Networked...

引用

International workshop on Robot Sensing (ROSE)

作者： Zuguang Liu Md Suruz Miah Department of Electrical and Computer Engineering Bradley University Peoria IL USA

ISBN: (数字)9798350362367

ISBN: (纸本)9798350362374

This paper presents a cooperative object transportation method using a group of autonomous networked cobots (collaborative robots). The cobot featured in this study is equipped with an elbow manipulator with three revolute joints, a wrist, and an end-effector for object manipulation. Additionally, it is mounted on a mobile platform base, propelled by omnidirectional wheels. Two instances of such a robot are assigned the task of synchronously coordinating in real-time to transport a large, delicate object, such as a sheet of glass, from one place to another, thereby representing a realistic industrial application. The proposed solution consists of three subsystems, namely manipulator control, twin navigation, and distributed autonomy. The elbow manipulator mounted on each cobot is controlled using iterative inverse kinematic method based on dynamic analysis. The robot's path planning strategy employs a leader-follower formation and utilizes a feedback control scheme to accurately track the payload's position. The autonomy of the robot is distributed such that the actuator commands for each cobot are determined and executed locally on each cobot's onboard system. For our study, we designed a scenario within CoppeliaSim, a commercial robot simulator, to mirror realworld industrial settings. The performance of the proposed cooperative object transportation method is evaluated within the scenario, leveraging the support of multiple physics engine for a comprehensive evaluation.

关键词： Wrist Service robots Robot kinematics Transportation Wheels Collaborative robots Robot sensing systems

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：