AI-Optimized Federated Learning with Dynamic Scheduling for High-Performance Computing

Federated Learning (FL) faces challenges in highperformance computing (HPC) due to bandwidth limitations and energy inefficiencies. This paper introduces an AI-driven dynamic scheduling framework leveraging reinforcement learning (RL) to optimize synchronization and energy usage. The proposed approach achieves over 99 % accuracy while improving energy efficiency by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{2 0 - 3 0 \%}$</tex>, significantly outperforming baseline methods in scalability and convergence speed. Comparative evaluations demonstrate its superiority in achieving low loss and adaptive scheduling under real-world constraints, paving the way for efficient distributed learning in HPC environments.