Research map / Reinforcement learning / Visual and representation RL
Swarm Reinforcement Learning for Adaptive Mesh Refinement
Niklas Freymuth, Philipp Dahlinger, Tobias Daniel Würth, Simon Reisch, Luise Kärger, Gerhard Neumann
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| combine per-element spatial reward with area scaling (ablation) | ASMR · Linear Elasticity (reward ablation) | normalized squared error vs number of elements | Improved |
| add area scaling term to a global (non-spatial) reward (ablation) | ASMR · Linear Elasticity (reward ablation) | normalized squared error vs number of elements | Hurt |
| replace mean-error reward with maximum-error reward (ablation) | ASMR · Linear Elasticity | refinement quality | No clear effect |
| use PPO instead of DQN (ablation) | ASMR and learned baselines · Poisson | mesh quality | Improved |
| add absolute positions to node features (ablation) | ASMR · Poisson | mesh quality | Hurt |
| add solution information and load function evaluations to node features (ablation) | ASMR · Poisson | mesh quality | Improved |
| increase training PDEs beyond 100 (ablation) | ASMR · Poisson | mesh quality | No clear effect |
| use MPN instead of GAT-like architecture (ablation) | VDGN · Poisson | mesh quality | Improved |
| use ASMR refinement instead of uniform mesh | Poisson (large spiral domains) · Poisson spiral domains | wall-clock speedup: more than factor 100 | Improved |
| use ASMR refinement instead of uniform mesh | all tasks · all tasks | wall-clock speedup: factor 2 to 30 | Improved |
| use ASMR instead of oracle/ZZ heuristics | mesh refinement · Stokes flow | error vs number of elements | Hurt |
| use ASMR instead of learned baselines at 6 refinement steps | mesh refinement · Laplace, 6 refinement steps | error vs number of elements | Improved |
Closest papers by meaning
- AMBER: Adaptive Mesh Generation by Iterative Mesh Resolution Prediction (NeurIPS 2025 · similarity 0.85)
- Model Swarms: Collaborative Search to Adapt LLM Experts via Swarm Intelligence (ICML 2025 · similarity 0.85)
- DesignX: Human-Competitive Algorithm Designer for Black-Box Optimization (NeurIPS 2025 · similarity 0.84)
- EvoMesh: Adaptive Physical Simulation with Hierarchical Graph Evolutions (ICML 2025 · similarity 0.84)
- MetaBox-v2: A Unified Benchmark Platform for Meta-Black-Box Optimization (NeurIPS 2025 · similarity 0.84)
- ConfigX: Modular Configuration for Evolutionary Algorithms via Multitask Reinforcement Learning (AAAI 2025 · similarity 0.83)
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.
