Graph network simulators can learn discontinuous, rigid contact dynamics.
Kelsey R. AllenTatiana Lopez-GuevaraYulia RubanovaKimberly StachenfeldAlvaro Sanchez-GonzalezPeter W. BattagliaTobias PfaffPublished in: CoRL (2022)
Keyphrases
- recurrent networks
- dynamic networks
- computer networks
- small world
- spanning tree
- graph theory
- dynamic model
- wireless sensor networks
- fully connected
- graph model
- graph structure
- communication networks
- sparsely connected
- three dimensional
- scale free
- graphical representation
- network structure
- graph representation
- complex networks
- network flow
- structured data
- path length
- random graphs
- weighted graph
- peer to peer
- sensor networks
- directed graph
- strongly connected
- degree distribution
- highly connected
- graph matching