Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Learning Neural Operators from Partial Observations via Latent Autoregressive Modeling
Jingren Hou, Hong Wang, Pengyu Xu, Chang Gao, Huafeng Liu, Liping Jing
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 mask-to-predict training (MPT) (ablation) | LANO · Navier-Stokes (point-wise 25%) | relative L2 error: 0.4958 -> 0.1274 | Improved |
| add mask-to-predict training (MPT) (ablation) | LANO · Diffusion-Reaction (patch-wise 50%, 200 samples) | relative L2 error: 0.6688 -> 0.1051 | Improved |
| add boundary-first propagation (ablation) | LANO · Navier-Stokes (point-wise 25%) | relative L2 error: 0.1963 -> 0.1274 | Improved |
| add boundary-first propagation (ablation) | LANO · Diffusion-Reaction (point-wise 25%) | relative L2 error: 0.4649 -> 0.0092 | Improved |
| add token mixer (ablation) | LANO · Navier-Stokes (point-wise 25%) | relative L2 error: 0.1459 -> 0.1274 | Improved |
| add token mixer (ablation) | LANO · Diffusion-Reaction (point-wise 25%) | relative L2 error: 0.0086 -> 0.0092 | No clear effect |
| replace MLP token mixer with attention (ablation) | LANO · Navier-Stokes (point-wise 25%) | relative L2 error: 0.1341 -> 0.1274 | Improved |
| replace MLP token mixer with attention (ablation) | LANO · Diffusion-Reaction (point-wise 25%) | relative L2 error: 0.0089 -> 0.0092 | No clear effect |
| increase latent features from 8 to 64 (ablation) | LANO · Navier-Stokes (point-wise 25%) | relative L2 error: 0.2244 -> 0.1198 | Improved |
| increase latent features from 8 to 64 (ablation) | LANO · Diffusion-Reaction (point-wise 25%) | relative L2 error: 0.0071 -> 0.0603 | Hurt |
| reuse encoder attention maps in decoder instead of recomputing (LANO vs LANO-S) (ablation) | LANO · Diffusion-Reaction (patch-wise 25%) | relative L2 error: 0.4718 -> 0.0275 | Improved |
| reuse encoder attention maps (LANO vs LANO-S) (ablation) | LANO · Navier-Stokes (patch-wise 25%) | relative L2 error: 0.1749 -> 0.1435 | Improved |
Closest papers by meaning
- Learning semilinear neural operators: A unified recursive framework for prediction and data assimilation. (ICLR 2024 · similarity 0.89)
- Stable Long-Horizon PDE Forecasting via Latent Structured Spectral Propagators (NeurIPS 2026 · similarity 0.88)
- Latent Neural Operator for Solving Forward and Inverse PDE Problems (NeurIPS 2024 · similarity 0.88)
- Data-Efficient Operator Learning via Unsupervised Pretraining and In-Context Learning (NeurIPS 2024 · similarity 0.88)
- CALM-PDE: Continuous and Adaptive Convolutions for Latent Space Modeling of Time-dependent PDEs (NeurIPS 2025 · similarity 0.87)
- Harnessing the Power of Neural Operators with Automatically Encoded Conservation Laws (ICML 2024 · similarity 0.87)
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.
