Corpus ID: 211252550

Learning to Simulate Complex Physics with Graph Networks

@article{SanchezGonzalez2020LearningTS,
  title={Learning to Simulate Complex Physics with Graph Networks},
  author={Alvaro Sanchez-Gonzalez and Jonathan Godwin and T. Pfaff and Rex Ying and J. Leskovec and P. Battaglia},
  journal={ArXiv},
  year={2020},
  volume={abs/2002.09405}
}
Here we present a machine learning framework and model implementation that can learn to simulate a wide 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… Expand

Figures from this paper

Scalable Graph Networks for Particle Simulations
Accurately Solving Physical Systems with Graph Learning
Data-Driven Continuum Dynamics via Transport-Teleport Duality
Learning Mesh-Based Simulation with Graph Networks
FORCENET: A GRAPH NEURAL NETWORK FOR LARGE-SCALE QUANTUM CHEMISTRY SIMULATION
  • 2020
Neural Relational Inference for Interacting Systems
VIGN: Variational Integrator Graph Networks
Machine learning–accelerated computational fluid dynamics
...
1
2
3
4
5
...

References

SHOWING 1-10 OF 44 REFERENCES
Graph networks as learnable physics engines for inference and control
Flexible Neural Representation for Physics Prediction
A Compositional Object-Based Approach to Learning Physical Dynamics
Latent Space Physics: Towards Learning the Temporal Evolution of Fluid Flow
Propagation Networks for Model-Based Control Under Partial Observation
Learning Particle Dynamics for Manipulating Rigid Bodies, Deformable Objects, and Fluids
Data-driven fluid simulations using regression forests
Hamiltonian Graph Networks with ODE Integrators
Learning to Control PDEs with Differentiable Physics
...
1
2
3
4
5
...