Research map / Graph learning / GNN architectures
Learning Graph Cellular Automata
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Learn the transition rule with a graph neural cellular automaton | Voronoi GCA at κ=0.42 · Voronoi GCA validation set | accuracy: 100% | Helped |
| Approximate the Boids transition rule with GNCA | Boids system · Boids trajectories | SampEn: 0.0369 -> 0.0302 | Mixed |
| Approximate the Boids transition rule with GNCA | Boids system · Boids trajectories | correlation dimension: 1.3125 -> 1.3017 | Mixed |
| Train for t=10 instead of using randomly sampled t∈[10,20] (ablation) | GNCA target convergence · Grid2d | minimum error: 1.19\cdot 10^{-6} -> 1.52\cdot 10^{-5} | Hurt |
| Train for t=10 instead of using randomly sampled t∈[10,20] (ablation) | GNCA target convergence · Bunny | minimum error: 9.76\cdot 10^{-5} -> 4.19\cdot 10^{-5} | Helped |
| Train for t=10 instead of using randomly sampled t∈[10,20] (ablation) | GNCA target convergence · Minnesota | minimum error: 1.29\cdot 10^{-6} -> 1.26\cdot 10^{-6} | Helped |
| Train for t=10 instead of using randomly sampled t∈[10,20] (ablation) | GNCA target convergence · Logo | minimum error: 2.49\cdot 10^{-4} -> 7.35\cdot 10^{-6} | Helped |
| Train for t=10 instead of using randomly sampled t∈[10,20] (ablation) | GNCA target convergence · SwissRoll | minimum error: 1.33\cdot 10^{-5} -> 1.09\cdot 10^{-5} | Helped |
| Train for t=20 instead of using randomly sampled t∈[10,20] (ablation) | GNCA target convergence · Grid2d | minimum error: 1.19\cdot 10^{-6} -> 3.66\cdot 10^{-6} | Hurt |
| Train for t=20 instead of randomly sampling t∈[10,20] (ablation) | GNCA target convergence · Bunny | minimum error: 9.76\cdot 10^{-5} -> 2.18\cdot 10^{-4} | Hurt |
| Train for t=20 instead of randomly sampling t∈[10,20] (ablation) | GNCA target convergence · Minnesota | minimum error: 1.29\cdot 10^{-6} -> 1.72\cdot 10^{-6} | Hurt |
| Train for t=20 instead of randomly sampling t∈[10,20] (ablation) | GNCA target convergence · Logo | minimum error: 2.49\cdot 10^{-4} -> 3.75\cdot 10^{-5} | Helped |
Closest papers by meaning
- Probing the Inductive Bias of Neural Networks through Learning Random Cellular Automata (ICML 2026 · similarity 0.86)
- CAX: Cellular Automata Accelerated in JAX (ICLR 2025 · similarity 0.85)
- 2D Spatial Reasoning with Adaptive Neural Cellular Automata (NeurIPS 2026 · similarity 0.85)
- Deep Neural Cellular Potts Models (ICML 2025 · similarity 0.83)
- Learning Mean Field Games on Sparse Graphs: A Hybrid Graphex Approach (ICLR 2024 · similarity 0.82)
- Integrating GNN and Neural ODEs for Estimating Non-Reciprocal Two-Body Interactions in Mixed-Species Collective Motion (NeurIPS 2024 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.