MeshGraphNets.jl - Adaptation of MeshGraphNets as NeuralODE

Explore the adaptation of MeshGraphNets as a NeuralODE in this conference talk from JuliaCon 2024. Dive into the world of simulating physical systems with spatially distributed domains using machine learning methods to improve computation time. Learn about the core concept of MeshGraphNets and its application in fluid dynamics and structural mechanics. Discover the design choices behind the Julia packages MeshGraphNets.jl and GraphNetCore.jl, including their implementation as NeuralODEs and compatibility with DifferentialEquations.jl. Understand how this approach allows for flexible training methods and optimized GPU execution. Examine the benefits of NeuralODE-based design through an industrial application in hydraulic brake systems.