AI Paper Spotlights - Multiscale Neural Operators for Solving Time-Independent PDEs
In this talk, our ML Researcher, Lisa Coiffard will be talking about Multiscale Neural Operators for Solving Time-Independent PDEs. Partial Differential Equations (PDEs) play a crucial role in various scientific and engineering tasks, from forecasting weather patterns to engineering design. In this talk, we will explore how we can solve these equations using modern, data-driven approaches focusing on Graph Neural Networks and Transformer architectures. Our discussion will highlight a novel graph rewiring technique, which helps overcome some common challenges in these methods, such as aggregating information across different scales. We will explore the practical implications of these methods with real-world applications, demonstrating how they have led to significant improvements in solving PDEs in diverse settings, from magnetic simulations in electric motors to flow dynamics in porous media.