Distributed Learning in wireless networks

The deployment of conventional centralized machine learning approaches in wireless networks requires wide bandwidth and large energy resources from the wireless users, and may potentially reveal the users’ privacy. Recently, the exponential increase of wireless devices leads to a large amount of distributed data is generated. By letting different types of wireless devices connected to the internet, different intelligent applications can be provided at the network edge. This project investigates distributed learning algorithms, which attempts to develop new distributed learning architectures for deploying distributed learning in heterogeneous wireless networks. We build on the foundations of federated learning and split learning paradigms to develop new hybrid split and federated learning, asynchronous federated learning algorithms. The application of these algorithms are planned for object tracking in wireless UAV networks, and user attention prediction in Metaverse over wireless networks.

Investigations initiated by developing hybrid federated and split learning approach for image recognition in wireless UAV networks, which is the foundation of dynamic object tracking relying on wireless networks. This research is mainly collaborated with King’s College London, Queen Mary University of London, The Hong Kong University of Science and Technology and Fudan university.

The current on-going research is to exploit personalized federated learning to predict user attention based on learning the multimodal data from each user, which is an important use for increasing immersive quality of experience of users in Metaverse. We are currently working on the initial framework and the foundations of training machine learning algorithms to predict user attention. This work is collaborating with Nanyang Technological University.

The project is aligned with the Artificial Intelligence, Machine Learning and Data Analytics research theme of the IDT.

For further information, please contact Dr. Xiaolan Liu

The following publications provide further insights into this research:

1. Xiaolan Liu, et al. “Wireless distributed learning: a new hybrid split and federated learning approach. " IEEE Transactions on Wireless Communications. (Accepted to be appear)
2. Xiaolan Liu, et al. “Energy Efficient User Scheduling for Hybrid Split and Federated Learning in Wireless UAV Networks. " IEEE ICC 2022
3. Xiaolan Liu, et al. “A Novel Hybrid Split and Federated Learning Architecture in Wireless UAV Networks. " IEEE ICC 2022
4. Xiaolan Liu, et al. “Distributed Intelligence in Wireless Networks. " IEEE Communications Surveys & Tutorials, (under review) 2022