GLOBECOM Best Journal Paper Award
ArtifiTialLeap, a research team of Ammar Kamal Abasi, Moayad Aloqaily, and Mohsen Guizani, has been awarded the Best Paper Award at IEEE GLOBECOM 2022. The team's "Grey Wolf Optimizer for Reducing Communication Cost of Federated Learning" paper was presented at the CSM Symposium.
Federated Learning (FL) is a type of Machine Learning (ML) technique that aims to maintain data security by only storing learned models on a server instead of server-side data. However, the communication between the server and clients in FL can be challenging, particularly when clients have limited bandwidth. ArtifiTialLeap's paper proposes a federated GWO (FedGWO) algorithm to reduce data communications, improving performance under unstable network conditions by transferring score principles rather than all client models' weights.
The team achieved a 13.55% average improvement in the global model's accuracy while decreasing the data capacity required for network communication. Furthermore, they demonstrated that FedGWO outperformed other methods, including FedAvg and Federated Particle Swarm Optimization (FedPSO), achieving a 5% reduction in accuracy loss when tested on unstable networks.
The IEEE GLOBECOM 2022 Best Paper Award is a prestigious recognition that acknowledges outstanding contributions in the field of communications. ArtifiTialLeap's research has significantly impacted federated learning, and its innovative approach can potentially improve data security and communication efficiency in this area. The team's achievement is a testament to their commitment to advancing research in the communications field and their dedication to producing high-quality work.