Research map / Reinforcement learning / Visual and representation RL
LASER: Learning Active Sensing for Continuum Field Reconstruction
Huayu Deng, Jinghui Zhong, Xiangming Zhu, Yunbo Wang, Xiaokang Yang
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 |
|---|---|---|---|
| add GRU history of length 3 vs none (ablation) | LASER latent dynamics · NS nu=1e-5 Out-t | rollout MSE (x1e-1): 5.38 -> 5.23 | Improved |
| increase history length from 3 to 5 (ablation) | LASER latent dynamics · NS nu=1e-5 Out-t | rollout MSE (x1e-1): 5.23 -> 7.88 | Hurt |
| add history on NS nu=1e-3 Out-t (ablation) | LASER latent dynamics · NS nu=1e-3 Out-t | rollout MSE (x1e-2): 0.822 -> 0.904 | Hurt |
| add dynamic quality filtering (ablation) | LASER · NS nu=1e-3, 64 sensors | recon MSE avg (x1e-3): 0.4521 -> 0.4339 | Improved |
| add dynamic quality filtering (ablation) | LASER · NS nu=1e-3, 256 sensors | recon MSE avg (x1e-3): 0.3913 -> 0.3024 | Improved |
| add sensor displacement (active sensing) over fixed layout (ablation) | LASER · NS nu=1e-3, 128 sensors | recon MSE avg (x1e-3): 0.8180 -> 0.3211 | Improved |
| lengthen reward horizon from 1 to 5 (ablation) | LASER · NS nu=1e-3, 256 sensors Out-t | recon MSE (x1e-3): 0.6136 -> 0.3380 | Improved |
| add iterative co-training of world model and policy | LASER · SST, 100 sensors | recon MSE: 0.6932 -> 0.6727 | Improved |
| use policy-induced sensors instead of random sensors (ablation) | LASER world model · NS nu=1e-3 Out-t | rollout MSE (x1e-2): 0.971 -> 0.498 | Improved |
| replace heuristic sensing (error-based) with learned policy (ablation) | LASER · NS nu=1e-3, 64 sensors | recon MSE (x1e-3): 26.61 -> 0.169 | Improved |
| replace PPO with proposed policy optimization (ablation) | LASER · NS nu=1e-3, 256 sensors | recon MSE avg (x1e-3): 0.304 -> 0.302 | Improved |
| replace PPO with proposed policy optimization (ablation) | LASER · Shallow-Water, 64 sensors | recon MSE avg (x1e-3): 0.456 -> 0.386 | Improved |
Closest papers by meaning
- StreamPhy: Streaming Inference of High-Dimensional Physical Dynamics via State Space Models (NeurIPS 2026 · similarity 0.85)
- Learning Dynamics under Environmental Constraints via Measurement-Induced Bundle Structures (ICML 2025 · similarity 0.84)
- Sparsely-Supervised Data Assimilation via Physics-Informed Schrödinger Bridge (NeurIPS 2026 · similarity 0.84)
- Christoffel-DPS: Optimal sensor placement in diffusion posterior sampling for arbitrary distributions (NeurIPS 2026 · similarity 0.83)
- Generating Full-field Evolution of Physical Dynamics from Irregular Sparse Observations (NeurIPS 2025 · similarity 0.83)
- Rapidly Adapting Policies to the Real-World via Simulation-Guided Fine-Tuning (ICLR 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.
