Project: Federated Learning on Embedded Systems 

Key information:

Abstract: 

Federated Learning is a distributed training strategy that enables different clients to collaboratively learn and apply a global model without having to share their data. It offers greater value in contrast to traditional methods such as ensuring data privacy and developing more robust models from training with vast quantities of decentralised data. Adapting Federated Learning on embedded systems can enhance the way we live, such as providing higher precision speech recognition by smart home systems and providing more reliable health and fitness statistics and diagnosis from data collected on motion-tracking wearables. Federated Learning on embedded devices presents a variety of challenges with regards to efficiency, scalability and heterogeneity. To this end, the project aims to develop solutions for federated learning to be efficient and scalable under resource-constrained environments.