Dynamic Middleware for Interoperable Federated Learning: Enabling Cross-Framework Communication
Mohammed AlKaldi, Abdullah AlShehri · 2024
Federated learning (FL) has emerged as a promising approach to machine learning, allowing multiple clients to collaborate on model training while maintaining data privacy. Despite its potential, the lack of interoperability between various FL frameworks poses a significant barrier to its widespread adoption, as it restricts the ability of diverse systems to communicate and share information effectively. This paper addresses this critical integration challenge by proposing a dynamic middleware layer designed to facilitate seamless communication and collaboration across different FL frameworks. By enabling cross-framework communication, our approach not only enhances the flexibility and scalability of FL systems but also improves overall model accuracy and efficiency. This dynamic middleware serves as a unifying interface that standardizes interactions between clients and servers, promoting more robust and efficient federated learning across heterogeneous environments.