在具有功率限制的无线联合学习网络中,联合客户端选择和CPU频率控制
Zhaohui Zhou1, Shijie Shi1, Fasong Wang1
1School of Electrical and Information Engineering, Zhengzhou University, Zhengzhou 450001, China.
Entropy (Basel, Switzerland)
|August 26, 2023
概括
本研究引入了一种优化的联合学习 (FL) 方法,以解决因资源异质性和非IID数据造成的缓慢培训时间. 新的算法共同优化客户端选择和CPU频率,以实现更快,更准确的机器学习.
科学领域:
- 机器学习 机器学习
- 分布式系统 分布式系统
- 无线网络 无线网络 无线网络
背景情况:
- 联合学习 (FL) 允许维护隐私的分布式培训,但面临来自异质资源的挑战,导致滞后者.
- 非独立且相同分布的 (非IID) 数据和资源限制进一步减缓了FL培训的速度.
- 现有的方法通常是单独优化客户端选择或资源分配,而不是联合.
研究的目的:
- 在无线联合学习中为客户端选择和计算能力 (CPU频率) 制定联合优化策略.
- 尽量减少一个成本函数,平衡学习延迟和非IID数据特征.
- 解决资源限制问题,提高联合学习的效率.
主要方法:
- 定义了一个包含学习延迟和非IID数据属性的成本函数.
- 制定了一个联合客户端选择和CPU频率控制问题,以尽量减少在功率限制下时间平均成本.
- 应用Lyapunov优化理论,将长期问题转化为连续的短期问题.
- 为客户端和边缘服务器开发了最佳客户端选择和CPU频率分配的算法.
主要成果:
- 拟议的算法实现了最佳的客户端选择和CPU频率控制.
- 理论分析为算法提供了性能保证.
- 与现有算法相比,模拟结果表明测试准确度更高.
- 这种方法有效地保持了低功耗.
结论:
- 客户端选择和CPU频率的联合优化对于无线网络中高效的联合学习至关重要.
- 拟议的算法有效地减轻了滞后者和非IID数据带来的挑战.
- 这种方法提供了一个有前途的解决方案,可以提高联合学习的性能和准确性,同时管理资源.
更多相关视频
07:34A Simple Stimulatory Device for Evoking Point-like Tactile Stimuli: A Searchlight for LFP to Spike Transitions
Published on: March 25, 2014
9.9K
10:36Author Spotlight: Optimization of Airflow Velocities in Battery Cooling Systems for Enhanced Thermal Performance and Reduced Energy Consumption
Published on: November 3, 2023
1.6K
相关概念视频
Load-frequency control
191
Load-frequency control (LFC) is vital for maintaining power system stability, ensuring that frequency and power flows remain within acceptable limits during load changes. Turbine-governor control eliminates rotor accelerations and decelerations following load changes. However, a steady-state frequency error persists when the change in the turbine-governor reference setting is zero. In an interconnected power system, each area agrees to export or import a scheduled amount of power through...
191
Maximum Power Transfer
287
Numerous practical applications within engineering disciplines, such as telecommunications, necessitate optimizing power delivery to a connected load. This pursuit, however, entails inherent internal losses, which can either equal or exceed the power supplied to the load. The Thevenin equivalent circuit is helpful in finding the maximum power a linear circuit can deliver to a load. It is assumed in this context that the load resistance can be adjusted.
By substituting the entire circuit with...
By substituting the entire circuit with...
287
Frequency-dependent Selection
22.1K
When the fitness of a trait is influenced by how common it is (i.e., its frequency) relative to different traits within a population, this is referred to as frequency-dependent selection. Frequency-dependent selection may occur between species or within a single species. This type of selection can either be positive—with more common phenotypes having higher fitness—or negative, with rarer phenotypes conferring increased fitness.
22.1K
Maximum Power Flow and Line Loadability
138
The maximum power flow for lossy transmission lines is derived using ABCD parameters in phasor form. These parameters create a matrix relationship between the sending-end and receiving-end voltages and currents, allowing the determination of the receiving-end current. This relationship facilitates calculating the complex power delivered to the receiving end, from which real and reactive power components are derived.
138
Distributed Loads: Problem Solving
668
Beams are structural elements commonly employed in engineering applications requiring different load-carrying capacities. The first step in analyzing a beam under a distributed load is to simplify the problem by dividing the load into smaller regions, which allows one to consider each region separately and calculate the magnitude of the equivalent resultant load acting on each portion of the beam. The magnitude of the equivalent resultant load for each region can be determined by calculating...
668
Fast Decoupled and DC Powerflow
233
The fast decoupled power flow method addresses contingencies in power system operations, such as generator outages or transmission line failures. This method provides quick power flow solutions, essential for real-time system adjustments. Fast decoupled power flow algorithms simplify the Jacobian matrix by neglecting certain elements, leading to two sets of decoupled equations:
233
