PhD Forum Abstract: Efficient Computing and Communication Paradigms for Federated Learning Data Streams

Abstract

In this work, we proposed an integration of Federated Learning with Apache Kafka, an open-source framework that enables the management of continuous data streams with fault tolerance, low latency, and horizontal scalability. Our main focus is to evaluate the impact of learning delays and network overhead when hundred of users are sending their model updates for the aggregation to improve the global model in Federated Learning.