Faster Client

Faster Client in Federated Learning (FL) focuses on optimizing the training process by mitigating the impact of slow clients, which significantly hinder overall training speed. Current research explores various client selection and scheduling strategies, including adaptive bandwidth allocation, dynamic tiering based on client capabilities, and availability-aware selection, aiming to minimize wall-clock training time without sacrificing model accuracy. These advancements address the heterogeneity in client resources and network conditions, improving the efficiency and scalability of FL for diverse applications. The resulting improvements in training speed have significant implications for deploying FL in resource-constrained environments and for accelerating the development of large-scale machine learning models.

Papers

February 15, 2024

Adaptive Federated Learning in Heterogeneous Wireless Networks with Independent Sampling
Jiaxiang Geng, Yanzhao Hou, Xiaofeng Tao, Juncheng Wang, Bing Luo
Sampling Technique Client Sampling Independent Sampling Adaptive Federated Learning Heterogeneous Wireless Network Faster Client

December 9, 2023

Speed Up Federated Learning in Heterogeneous Environment: A Dynamic Tiering Approach
Seyed Mahmoud Sajjadi Mohammadabadi, Syed Zawad, Feng Yan, Lei Yang
Speed Effect Heterogeneous Computing Dynamic Scheduling Heterogeneous Environment Faster Client

November 25, 2022

MDA: Availability-Aware Federated Learning Client Selection
Amin Eslami Abyane, Steve Drew, Hadi Hemmati
Client Selection Federated Learning Client Faster Client

October 12, 2022

Aergia: Leveraging Heterogeneity in Federated Learning Systems
Bart Cox, Lydia Y. Chen, Jérémie Decouchant
Deep Learning Federated Learning Performance Bottleneck Faster Client

July 10, 2022

FedSS: Federated Learning with Smart Selection of clients
Ammar Tahir, Yongzhou Chen, Prashanti Nilayam
Client Selection Selection Policy Client Collaboration Faster Client

Faster Client

Papers

Adaptive Federated Learning in Heterogeneous Wireless Networks with Independent Sampling

Speed Up Federated Learning in Heterogeneous Environment: A Dynamic Tiering Approach

MDA: Availability-Aware Federated Learning Client Selection

Aergia: Leveraging Heterogeneity in Federated Learning Systems

FedSS: Federated Learning with Smart Selection of clients