Improving the Federated Aggregation Communication Performance in xFFL

Improve communication efficiency in xFFL, an AI and federated learning training framework, with particular emphasis on the aggregation phase.

The candidate will: 1. Study the architecture and communication mechanisms of xFFL. 2. Profile the aggregation phase to identify performance bottlenecks. 3. Review relevant communication optimization techniques. 4. Design and implement one or more improvements. 5. Benchmark the revised implementation under representative workloads. 6. Compare the optimized implementation with the original version.

3 to 9 months according to the student's background.

• Good knowledge of Python
• Knowledge of distributed systems

1. Gianluca Mittone, & Marco Aldinucci (2026). Accelerating Sharded Data Parallelism at Scale with Federated Learning. In Euro-Par 2026: Parallel Processing - 32nd European Conference on Parallel and Distributed Processing, Pisa, Italy, August 24-28, 2026, Proceedings, Part II (pp. 437–451). Springer.