Abstract:With rapid advancement of Internet of Things (IoT) and social networking applications generating large amounts of data at or close to the network edge, Mobile Edge Computing (MEC) has naturally been proposed to bring model training closer to where data is produced. However, there still exists privacy concern since typical MEC frameworks need to transfer sensitive data resources from data collection end devices/clients to MEC server. So the concept of Federated Learning (FL) has been introduced which supports privacy-preserved collaborative machine learning involving multiple clients coordinated by the MEC server without centralizing the private data. Unfortunately, FL is prone to multiple challenges: 1) systems heterogeneity between clients causes straggler issue, and 2) statistical heterogeneity between clients brings about objective inconsistency problem, both of which may lead to a significant slow-down in the convergence speed in heterogeneous MEC environment. In this paper, we propose a novel framework, FedAda (Federated Adaptive Training), that incorporates systems capabilities and data characteristics of the clients to adaptively assign appropriate workload to each client. The key idea is that instead of running a fixed number of local training iterations as in Federated Averaging (FedAvg), our algorithm adopts an adaptive workload assignment strategy by minimizing the runtime gap between clients and maximizing convergence gain in heterogeneous MEC environment. Moreover, we design a light mechanism extending FedAda to accelerate the convergence speed by further fine-tuning the workload assignment based on the global convergence status in each communication round. We evaluate FedAda on CIFAR-10 dataset to explore the performance of the algorithm in the simulated heterogeneous MEC environment. Experimental results show that FedAda is able to assign appropriate amount of workload to each client and substantially reduces the convergence time by up to 49.5% compared to FedAvg in heterogeneous MEC environment. In addition, we demonstrate that fine-tuning the workload assignment can help FedAda improve the learning performance in heterogeneous mobile edge computing environment.

FAST: Enhancing Federated Learning Through Adaptive Data Sampling and Local Training

FedSpeed: Larger Local Interval, Less Communication Round, and Higher Generalization Accuracy

Fast Federated Learning in the Presence of Arbitrary Device Unavailability

Efficient Federated Learning via Local Adaptive Amended Optimizer with Linear Speedup

Enhancing Federated Learning Convergence with Dynamic Data Queue and Data Entropy-driven Participant Selection

AsyFed: Accelerated Federated Learning with Asynchronous Communication Mechanism

FLAS: Computation and Communication Efficient Federated Learning via Adaptive Sampling

FedBalancer: Data and Pace Control for Efficient Federated Learning on Heterogeneous Clients

FedPD: A Federated Learning Framework with Optimal Rates and Adaptivity to Non-IID Data.

Enhancing Convergence in Federated Learning: A Contribution-Aware Asynchronous Approach

Asynchronous Online Federated Learning for Edge Devices with Non-IID Data

FedCos: A Scene-adaptive Federated Optimization Enhancement for Performance Improvement

FedLion: Faster Adaptive Federated Optimization with Fewer Communication

An Efficient Down Sampling Scheme Based on a Matching Factor for Federated Learning

Accelerating Federated Learning by Selecting Beneficial Herd of Local Gradients

Automated Federated Learning in Mobile Edge Networks -- Fast Adaptation and Convergence

Device Scheduling with Fast Convergence for Wireless Federated Learning

Sample-level Data Selection for Federated Learning.

FedSW: Federated learning with adaptive sample weights

FedAda: Fast-convergent adaptive federated learning in heterogeneous mobile edge computing environment

FedLC: Accelerating Asynchronous Federated Learning in Edge Computing