Tobias Pfaff (DeepMind): Learning to Simulate Complex Physics with Graph Networks

We have invited Tobias Pfaff from DeepMind to speak about his team's recent paper which presents a general framework called "Graph Network-based Simulators (GNS)" for learning physical simulation with a single model. It achieves state-of-the-art performance across a variety of challenging physical domains, involving fluids, rigid solids, and deformable materials interacting with one another.

For slides and more information on the paper, visit https://ai.science/e/learning-to-simulate-complex-physics-with-graph-networks--reO5bMSsNtcJcpr1Muvt

Speaker / coauthor: Tobias Pfaff (Research Scientist at DeepMind)

Abstract

Here we present a general framework for learning simulation, and provide a single model implementation that yields state-of-the-art performance across a variety of challenging physical domains, involving fluids, rigid solids, and deformable materials interacting with one another. Our framework---which we term "Graph Network-based Simulators" (GNS)---represents the state of a physical system with particles, expressed as nodes in a graph, and computes dynamics via learned message-passing. Our results show that our model can generalize from single-timestep predictions with thousands of particles during training, to different initial conditions, thousands of timesteps, and at least an order of magnitude more particles at test time. Our model was robust to hyperparameter choices across various evaluation metrics: the main determinants of long-term performance were the number of message-passing steps, and mitigating the accumulation of error by corrupting the training data with noise. Our GNS framework is the most accurate general-purpose learned physics simulator to date, and holds promise for solving a wide range of complex forward and inverse problems.
AISC Spotlight

Видео Tobias Pfaff (DeepMind): Learning to Simulate Complex Physics with Graph Networks канала ML Explained - Aggregate Intellect - AI.SCIENCE