A Paper on Learning to Simulate Complex Physics

A paper called Learning to Simulate Complex Physics with Graph Networks discusses a general framework for learning simulation which achieves state-of-the-art performance across fluids, rigid solids, and deformable materials interacting with one another.

First, have a look at a video by Two Minute Papers:

"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," states the team. The new framework is said to capable of dealing with thousands of particles during training adapting to different initial conditions.

The team states that this model is the most accurate general-purpose learned physics simulator to date that can deal with a wide range of complex tasks. You can find the full paper here.