Scalable federated learning via memory-efficient and concurrent aggregation.
Jan 1, 2026
SRFL targets scalable and resilient federated learning systems across heterogeneous compute and network environments. The project includes: FedDES, a discrete-event based performance simulation framework for federated learning systems. FedMECA, a memory-efficient and concurrent aggregation approach for scalable federated learning. Long-haul RDMA studies for geo-distributed federated learning, including simulation, modeling, and real-world testbed validation.
Jan 1, 2025
This project targets HPC-AI convergence for efficient large-scale machine learning, including scheduling, optimization, characterization, and fault-tolerant training systems. The project includes: HPC-R1, a characterization of inference and distillation performance for large reasoning models on HPC-scale GPU clusters and interconnects. SPARe, a fault-tolerant LLM pretraining system for 100k+ GPU scale using stacked parallelism and adaptive reordering.
Jan 1, 2025