Optimizing Federated Learning Approach: Literature Survey and Open Points

Paper ID : 1024-ICEEM2023 (R1)

Authors

Tharwat Elsayed Ismail Abdalla *1, Abdalla Nabil2, Mohamed M Elrashidy2, Ayman EL-SAYED3 141511 2Menoufia university 3Menoufia University

Abstract

Federated Learning (FL) is a decentralized machine learning strategy that ensures training data stays on personal devices while also facilitating collaborative machine learning of complicated models among dispersed devices. FL is used to circumvent Mobile Edge Computing’s constraints (MEC). In FL, mobile devices utilize their local data to train an ML model that is controlled by the FL server. The model changes, i.e. the model’s weights, are subsequently sent to the FL server for collection. This stage is performed several times until the desired precision is obtained. FL has several issues, including communication costs, resource allocation, privacy, and security. The Federated Learning method is outlined in this work, and existing solutions to overcome its constraints are categorized and reviewed. Furthermore, there are several research directions and unanswered questions that must be addressed while considering a better method to Maximize Federated Learning Models in The Mobile Edge Computing Environment.

Keywords

Artificial Intelligence; Mobile Edge Computing; Machine Learning; Federated Learning; Data privacy; Blockchain

Status: Accepted