Research map / Probabilistic and scientific modeling / Dynamical systems
PETAL: Physics Emulation Through Averaged Linearizations for Solving Inverse Problems
Jihui Jin, Etienne Ollivier, Richard Touret, Matthew McKinley, Karim Sabra, Justin Romberg
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 |
|---|---|---|---|
| use weighted average of linearizations instead of plain average (WA-LFM vs A-LFM) (ablation) | A-LFM · Ocean acoustic tomography SSP inversion | RMSE (m/s): 0.585 -> 0.577 | Improved |
| add learned SSP subspace to weighted average (WA-LFM + Dec) (ablation) | WA-LFM · Ocean acoustic tomography SSP inversion | RMSE (m/s): 0.577 -> 0.508 | Improved |
| add learned arrival-time transform (WAN vs WA-LFM) (ablation) | WA-LFM · Ocean acoustic tomography SSP inversion | RMSE (m/s): 0.577 -> 0.372 | Improved |
| add SSP subspace on top of weighted average plus AT transform (full PETAL vs WAN) (ablation) | WAN · Ocean acoustic tomography SSP inversion | RMSE (m/s): 0.372 -> 0.343 | Improved |
| use many reference linearizations (A-LFM) instead of a single reference (LFM) | LFM · Ocean acoustic tomography SSP inversion | RMSE (m/s): 0.608 -> 0.585 | Improved |
| initialize PETAL with LFM solution instead of average SSP | PETAL · Ocean acoustic tomography SSP inversion | RMSE (m/s): 0.343 -> 0.342 | No clear effect |
| initialize PETAL with Tik solution instead of average SSP | PETAL · Ocean acoustic tomography SSP inversion | RMSE (m/s): 0.343 -> 0.339 | No clear effect |
| initialize LFM with LFM solution (further optimization) | LFM · Ocean acoustic tomography SSP inversion | RMSE (m/s): 0.608 -> 0.625 | Hurt |
| initialize LFM with Tik solution | LFM · Ocean acoustic tomography SSP inversion | RMSE (m/s): 0.608 -> 0.602 | No clear effect |
| initialize MLP with LFM solution | MLP · Ocean acoustic tomography SSP inversion | RMSE (m/s): 0.398 -> 0.402 | No clear effect |
| PETAL vs MLP surrogate (physics-informed linearizations) | MLP · Ocean acoustic tomography SSP inversion | RMSE (m/s): 0.398 -> 0.343 | Improved |
| train on slices 1-9 and test on unseen slice 10 test-time split | PETAL · Unseen slice 10 | RMSE (m/s): 0.149 -> 0.337 | Hurt |
Closest papers by meaning
- OceanCBM: A Concept Bottleneck Model for Mechanistic Interpretability in Ocean Forecasting (NeurIPS 2026 · similarity 0.85)
- Why Learning Rediscovers the Closed-Form Diagonal Regularizer (NeurIPS 2026 · similarity 0.85)
- A Unified Framework for Forward and Inverse Problems in Subsurface Imaging using Latent Space Translations (ICLR 2025 · similarity 0.85)
- Geometric and Physical Constraints Synergistically Enhance Neural PDE Surrogates (ICML 2025 · similarity 0.84)
- Multiple Physics Pretraining for Spatiotemporal Surrogate Models (NeurIPS 2024 · similarity 0.83)
- Sparsely-Supervised Data Assimilation via Physics-Informed Schrödinger Bridge (NeurIPS 2026 · 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.
