检索结果-内蒙古大学图书馆

adaptive client and communication optimizations in Federated Learning

INFORMATION SYSTEMS 2023年 116卷

作者： Wu, Jiagao Wang, Yu Shen, Zhangchi Liu, Linfeng Nanjing Univ Posts & Telecommun Sch Comp Nanjing 210023 Peoples R China Jiangsu Key Lab Big Data Secur & Intelligent Proc Nanjing 210023 Peoples R China

Federated Learning (FL) is a very effective distributed machine learning framework that enables a large number of devices to jointly train models without sharing raw data. However, the process of iterative learning and data communication in FL can be time-consuming, depending heavily on the number of clients participating in training and the number of local iterations between two consecutive global aggregations (communication period). In this paper, we analyze the runtime of the FL framework and propose a joint optimization problem, which considers both the number of local clients involved in global training and the communication period to minimize the error-runtime convergence of FedAvg. By theoretical analysis, we obtain the optimal solutions of this optimization problem in closed forms under the assumption of different probability distributions of the client's local computation time. In addition, we design an adaptive control algorithm that can dynamically select the number of clients and communication period during FL training to speed up the convergence of the model. Experimental results validate our theoretical analysis and show that the proposed algorithm has outstanding performance compared with other related FL control algorithms.(c) 2023 Elsevier Ltd. All rights reserved.

关键词： Federated learning Distributed machine learning Optimization problem Theoretical analysis adaptive control algorithm Edge computing

来源：评论

学校读者我要写书评

暂无评论

Feedback control for deep brain stimulation for motor disorders

引用

HEALTHCARE TECHNOLOGY LETTERS 2020年第3期7卷 72-75页

作者： McDermott, Hugh J. Sinclair, Nicholas C. Bion Inst East Melbourne Vic 3002 Australia Univ Melbourne Med Bion Dept East Melbourne Vic Australia

Deep brain stimulation (DBS) is an effective treatment for many people living with Parkinson's disease (PD). Although the primary treatment for PD is based on medications, disease progression eventually leads to inadequate symptom control. DBS provides benefits by alleviating motor dysfunctions such as muscle rigidity and tremor. DBS devices deliver electric pulse trains into specific brain regions via implanted electrodes. Existing DBS systems usually provide continuous stimulation with constant settings of parameters such as the amount of charge delivered per pulse. However, PD is characterised by fluctuations in the severity and frequency of impairments. DBS would be improved if stimulation settings were adjusted automatically in response to each patient's ever-changing needs. This requires a device incorporating sensing of signals that estimate the severity of motor impairment linked to an adaptive control algorithm that optimises therapeutic stimulation. Several types of signals are candidates for this function. Spontaneous local field potentials recorded by the DBS electrodes have shown promise in some experimental studies of adaptive DBS. More recently, DBS-evoked potentials have been reported. In particular, evoked resonant neural activity has properties including a larger amplitude than spontaneous potentials, suggesting it may be a suitable feedback signal to control adaptive DBS.

关键词： diseases biomedical electrodes feedback prosthetics brain medical disorders patient treatment microelectrodes bioelectric potentials muscle neurophysiology deep brain stimulation motor disorders Parkinson disease PD primary treatment disease progression inadequate symptom control motor dysfunctions muscle rigidity tremor DBS devices electric pulse trains specific brain regions DBS systems continuous stimulation motor impairment adaptive control algorithm therapeutic stimulation DBS electrodes adaptive DBS DBS-evoked potentials feedback signal feedback control

来源：评论

学校读者我要写书评

暂无评论

adaptive wire diameter control of the high-precision optical fibre drawing machine

引用

OPTICAL FIBER TECHNOLOGY 2021年 63卷 102462-102462页

作者： Cao, Yang Qian, Yunsheng Zhang, Jiawei Hao, Jinguang Wang, Honggang Yang, Minjie Nanjing Univ Sci & Technol Coll Elect & Opt Engn Nanjing 210094 Peoples R China North Night Vis Technol Co Ltd Nanjing 210094 Peoples R China Ludong Univ Coll Informat & Elect Engn Yantai 264025 Peoples R China

Different from the conventional methods such as the Proportional-Integral-Derivative (PID)-based method, an adaptive control algorithm (ACA) is proposed in this paper to control the wire diameter of high-precision fibre drawing machines (FDMs). A simple FDM model is established to acquire the relationship between fibre diameter and drawing velocity before the ACA is introduced. The comparison simulation of the accuracy of fibre diameter between the proposed ACA and traditional PID is made, and the ACA is applied to actual fibre drawing processes, so as to validate the ACA. The comparative results show that the fluctuation of filament diameter is smaller and the accuracy is higher under the ACA control. In actual fibre filament drawing processes, the ACA can limit the diameter error within ?2?m, which meets the production requirements.

关键词： Fibre drawing machine Wire diameter control adaptive control algorithm PID control

来源：评论

学校读者我要写书评

暂无评论

adaptive Least Mean Square controller for Power Quality Enhancement in Solar Photovoltaic System

引用

ENERGIES 2022年第23期15卷 8909页

作者： Karchi, Nalini Kulkarni, Deepak Perez de Prado, Rocio Divakarachari, Parameshachari Bidare Patil, Sujata N. Desai, Veena Gogte Inst Technol Dept Elect & Elect Engn Belagavi 590006 India KLE Dr MS Sheshgiri Coll Engn & Technol Dept Elect & Elect Engn Belagavi 590008 India Univ Jaen Telecommun Engn Dept Jaen 23700 Spain Nitte Meenakshi Inst Technol Dept Elect & Commun Engn Bangalore 560064 India KLE Dr MS Sheshgiri Coll Engn & Technol Dept Elect & Commun Engn Belagavi 590008 India Gogte Inst Technol Dept Elect & Commun Engn Belagavi 590006 India

The objective of the proposed work is to develop a Maximum Power Point Tracking (MPPT) controller and inverter controller by applying the adaptive least mean square (LMS) algorithm to control the total harmonics distortion of a solar photovoltaic system. The advantage of the adaptive LMS algorithm is given by its simplicity and reduced required computational time. The adaptive LMS algorithm is applied to modify the Perturb and Observe (P&O), MPPT controller. In this controller, the adaptive LMS algorithm is used to predict solar photovoltaic power. The adaptive LMS maximum power point tracking controller gives better optimal solutions with less steady error 0.7% (6 watts) and 0% peak overshot in power with the tradeoff being more settling time at 0.33 s. The development of the inverter control law is performed using the d-q frame theory. This helps to reduce the number of equations to build a control law. The load current, grid current and grid voltage are sensed and transformed into d and q components. This adaptive LMS control law is used to extract the reference grid currents and, later, to compare them with the actual grid currents. The result of this comparison is used to generate the switching gate pulses for the inverter switches. The proposed controllers are developed and implemented with a solar PV system in MATLAB Simulink. The total harmonics distortion in grid and load current (3.25% and 7%) and voltage (0%) is investigated under linear and non-linear load conditions with changes in solar irradiations. The analysis is performed by selecting step incremental values and sampling time.

关键词： adaptive control algorithm inverter controller least mean square maximum power point tracking photovoltaic system power quality issues

来源：评论

学校读者我要写书评

暂无评论

Robust Adaptation in Dynamically Switching Load Frequency control 21st

Robust Adaptation in Dynamically Switching Load Frequency Co...

引用

21st IFAC World Congress on Automatic control - Meeting Societal Challenges

作者： Tao, Tian Roy, Spandan Yuan, Shuai Baldi, Simone Southeast Univ Sch Math Nanjing Peoples R China Southeast Univ Sch Cyber Sci & Engn Nanjing Peoples R China Delft Univ Technol Delft Ctr Syst & Control Delft Netherlands Harbin Inst Technol Sch Astronaut Harbin Heilongjiang Peoples R China Int Inst Informat Technol Hyderabad Robot Res Ctr Hyderabad Telangana India

In recent years, heuristics for adaptive solutions to load frequency control (LFC) in power systems have been proposed that include adapting the LFC targets or adapting the participation factor for the resources. However, stability guarantees for these adaptation ideas are missing, especially in the presence of switching/evolving topologies of the power system. In today's smart grids, switching topologies often arise from reconfiguration and resilience against faults or from switching among different control areas in order to dampen oscillations and face cyber attacks. This work proposes a novel LFC framework in which adaptation and switching topologies are combined in a provably stable way. Copyright (C) 2020 The Authors.

关键词： Power systems load frequency control adaptive control algorithm grid resilience switching/evolving topologies

来源：评论

学校读者我要写书评

暂无评论

adaptive control technique for suppression of resonance in grid-connected PV inverters

引用

IET POWER ELECTRONICS 2019年第6期12卷 1479-1486页

作者： Goh, Hongsoo Armstrong, Matthew Zahawi, Bashar Northern Powergrid Protect & Tech Serv Castleford WF10 SDS England Newcastle Univ Sch Engn Newcastle Upon Tyne NE1 7RU Tyne & Wear England Khalifa Univ Dept Elect & Comp Engn Abu Dhabi 127788 U Arab Emirates

Grid operating conditions have a significant effect on the harmonic and resonant performance of grid-connected photovoltaic (PV) inverters and changes in grid impedance can cause a notable change in the resonant excitation between the PV inverter and the grid. This study proposes an adaptive control algorithm for grid-connected PV inverters to suppress the resonance condition excited by grid inductance variation, resulting from the dynamic changes in the operating conditions of the distribution network. The causes of resonance between grid-connected PV inverters and the distribution grid are discussed and the design of an active band-pass filter for capturing resonance is described. The proportional gain within the proportional-integral controller is then adaptively controlled in real time to compensate for changes in the grid impedance and suppress resonant excitation while maintaining excellent low-order harmonic performance compared with alternative fixed gain controller techniques, particularly for systems with high values of grid inductance. The performance of the proposed controller is experimentally verified using a 240 V, 2 kVA single-phase grid-connected inverter.

关键词： photovoltaic power systems invertors adaptive control band-pass filters power grids adaptive control technique grid-connected PV inverters grid operating conditions harmonic performance resonant performance grid-connected photovoltaic inverters grid impedance PV inverter adaptive control algorithm resonance condition grid inductance variation distribution grid single-phase grid-connected inverter resonant excitation suppression active band-pass filter voltage 240 0 V apparent power 2 0 kVA

来源：评论

学校读者我要写书评

暂无评论

Embedded and Enterprise Fault Management for Aerospace Domain

Embedded and Enterprise Fault Management for Aerospace Domai...

引用

AIAA Science and Technology Forum and Exposition (AIAA SciTech)

作者： Gogg, Thomas Cheah, Joseph Ko, Gene M. Lee, Tony Kapadia, Ravi Venkatesh, Meera Gen Atom Electromagnet Syst Grp Intelligent Automat & Analyt GA EMS San Diego CA 92127 USA

ISBN: (数字)9781624105951

ISBN: (纸本)9781624105951

Prognostics and health management (PHM) facilitates the ability to readily forecast the likelihood of component failures within a system. This can provide significant cost savings with respect to system availability, maintenance costs, and in some cases, the saving of human lives. A prototypical PHM application intended to support aircraft with hybrid electric distributed propulsion is being presented. Such an aircraft may contain multiple electric motors each equipped with independent control and propulsion systems. Embedded PHM is necessary for real-time monitoring, assessing, and reporting of system health to the pilot. It also supports advanced adaptive control algorithms. An enterprise PHM system is used to store, process, and analyze the historical sensor data for comprehensive fault model processing, root cause analysis, and usage monitoring. Fault prediction models developed on the enterprise system can be deployed to the embedded computing device for real-time monitoring of critical components.

关键词： Onboard Sensors Embedded Computing Electric Motors adaptive control algorithm Propulsion System Pilot Naive Bayes Classifier Graphical User Interface Electric Current Edge Computing

来源：评论

学校读者我要写书评

暂无评论

Closed-Loop control of a Piezo-Fluidic Amplifier

引用

AIAA JOURNAL 2020年第6期58卷 2414-2427页

作者： Nicholls, Chris Bacic, Marko Univ Oxford Oxford Thermofluids Inst Oxford OX2 0ES England Rolls Royce PLC London England

Fluidic valves based on the Coanda effect are increasingly being considered for use in aerodynamic flow control applications. A limiting factor is their variation in switching time, which often precludes their use. The purpose of this paper was to demonstrate the closed-loop control of a recently developed novel piezo-fluidic valve that reduces response time uncertainty at the expense of operating bandwidth. Use is made of the fact that a fluidic jet responds to a piezo tone by deflecting away from its steady-state position. A control signal used to vary this deflection is amplitude modulated onto the piezo tone. Using only a pressure measurement from one of the device output channels, an output-based linear quadratic Gaussian regulator was designed to follow a desired reference deflection, achieving control of a 90 m.s(-1) jet. Finally, the controller's performance in terms of disturbance rejection and response time predictability is demonstrated.

关键词： adaptive control algorithm Coanda Effect Linear Quadratic Gaussian Aerodynamic Flows Power Spectral Density Shear Layers Mass Flow Rate Piezoelectric Transducers Dynamic Response Extended Kalman Filter

来源：评论

学校读者我要写书评

暂无评论

Robust Adaptation in Dynamically Switching Load Frequency control ⁎

引用

IFAC-PapersOnLine 2020年第2期53卷 13460-13465页

作者： Tian Tao Spandan Roy Shuai Yuan Simone Baldi Delft Center for Systems and Control TU Delft The Netherlands Robotics Research Centre International Institute of Information Technology Hyderabad India School of Astronautics Harbin Institute of Technology China School of Mathematics Southeast University Nanjing China School of Cyber Science and Engineering Southeast University Nanjing China

In recent years, heuristics for adaptive solutions to load frequency control (LFC) in power systems have been proposed that include adapting the LFC targets or adapting the participation factor for the resources. However, stability guarantees for these adaptation ideas are missing, especially in the presence of switching/evolving topologies of the power system. In today’s smart grids, switching topologies often arise from reconfiguration and resilience against faults or from switching among different control areas in order to dampen oscillations and face cyber attacks. This work proposes a novel LFC framework in which adaptation and switching topologies are combined in a provably stable way.

关键词： Power systems load frequency control adaptive control algorithm grid resilience switching evolving topologies

来源：评论

学校读者我要写书评

暂无评论

Closed-Loop control in Active Target Defense Using Machine Learning

Closed-Loop Control in Active Target Defense Using Machine L...

引用

AIAA Science and Technology (SciTech) Forum and Exposition

作者： Lau, Mackenzie Steffens, Michael Mavris, Dimitri Georgia Inst Technol Aerosp Syst Design Lab Atlanta GA 30332 USA Sch Aerosp Engn Atlanta GA 30332 USA

ISBN: (数字)9781624105784

ISBN: (纸本)9781624105784

Capturing the impact of human decision making processes in engineering design presents a significant challenge to engineers. However, humans have repeatedly demonstrated the capacity to overcome technological disadvantage through the use of novel tactics. This indicates that not accounting for tactics can result in designs which underperform. The Active Target Defense scenario is a well-studied problem and has been solved through the lens of optimal control theory. It also posses many elements of more challenging problems, such as those involving human decision making processes. This makes the problem a good candidate for exploring the applicability of new techniques to accommodate those elements. In particular, the use of machine learning has proven to be an effective means to achieve super-human performance in certain scenarios. This work applied machine learning to the Active Target Defense game and leveraged existing literature to validate the method. The results show good, though not perfect, agreement with optimal solutions and lay a foundation for further investigations into this class of problems.

关键词： Reinforcement Learning adaptive control algorithm Optimal control Theory Proportional Navigation Heading Angle Guidance Laws Neural Networks Air to Air Missiles United States Department of Defense Guidance algorithms

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：