Our work introduces a GNN-based method for learning SPAI preconditioners, enabling efficient and robust training across various datasets and outperforming traditional and learning-based approaches in accelerating Conjugate Gradient solvers on GPUs.

We propose Differential RL, a physics-informed framework that reformulates RL as a differential control problem. Its algorithm, dfPO, achieves pointwise convergence and outperforms standard RL in low-data scientific computing tasks.