Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Unsupervised Learning of Full-Waveform Inversion: Connecting CNN and Partial Differential Equation in a Loop
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 |
|---|---|---|---|
| Increase unlabeled training data from 24K to 48K | UPFWI · FlatFault | MSE: 1705.35 -> 1146.09 | Helped |
| Increase unlabeled training data from 24K to 48K | UPFWI · CurvedFault | MSE: 5712.25 -> 3639.96 | Helped |
| Add pixel-ℓ1 loss to pixel-ℓ2 loss (ablation) | UPFWI · FlatFault | velocity MSE: 10014.47 -> 2999.55 | Helped |
| Add perceptual loss to pixel-ℓ1+ℓ2 loss (ablation) | UPFWI · FlatFault | velocity MSE: 2999.55 -> 1705.35 | Helped |
| Replace CNN with ViT (ablation) | UPFWI · FlatFault | MSE: 1705.35 -> 11029.01 | Hurt |
| Replace CNN with MLP-Mixer (ablation) | UPFWI · FlatFault | MSE: 1705.35 -> 4177.37 | Hurt |
| Use unsupervised UPFWI rather than supervised InversionNet | FWI · Marmousi | MSE: 45936.23 -> 125825.75 | Hurt |
| Use unsupervised UPFWI rather than supervised InversionNet | FWI · Salt | MSE: 8669.98 -> 164595.28 | Hurt |
| Increase Gaussian noise level | UPFWI · FlatFault | MSE: 1343.21; 4010.78; 7592.57 | Mixed |
| Increase missing-trace rate | UPFWI · FlatFault | MSE: 1772.05; 3504.25; 16731.69 | Hurt |
Closest papers by meaning
- A Unified Framework for Forward and Inverse Problems in Subsurface Imaging using Latent Space Translations (ICLR 2025 · similarity 0.92)
- Auto-Linear Phenomenon in Subsurface Imaging (ICML 2024 · similarity 0.87)
- A Novel Diffusion Model for Pairwise Geoscience Data Generation with Unbalanced Training Dataset (AAAI 2025 · similarity 0.84)
- CALM-PDE: Continuous and Adaptive Convolutions for Latent Space Modeling of Time-dependent PDEs (NeurIPS 2025 · similarity 0.84)
- Physics-Informed Deep Inverse Operator Networks for Solving PDE Inverse Problems (ICLR 2025 · similarity 0.83)
- Symmetric Basis Convolutions for Learning Lagrangian Fluid Mechanics (ICLR 2024 · 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.