Research map / Learning theory and algorithms / Combinatorial and search optimization
Benchmarking Deep Inverse Models over time, and the Neural-Adjoint method
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 |
|---|---|---|---|
| Use Neural Adjoint instead of the strongest listed baseline, cVAE | Inverse modeling · Ballistics (D1) | r_T=200 MSE: 2.80e-7 -> 2.50e-7 | Helped |
| Use Neural Adjoint instead of the strongest listed baseline, cINN | Inverse modeling · Sine wave (D2) | r_T=200 MSE: 2.72e-5 -> 1.33e-7 | Helped |
| Use Neural Adjoint instead of the strongest listed baseline, cINN | Inverse modeling · Robotic arm (D3) | r_T=200 MSE: 8.80e-7 -> 6.61e-7 | Helped |
| Use Neural Adjoint instead of the strongest listed baseline, cINN | Inverse modeling · Meta-material (D4) | r_T=200 MSE: 4.45e-4 -> 6.67e-5 | Helped |
| Add boundary loss with uniform initialization (ablation) | Neural Adjoint (NA), Ballistics · Ballistics (D1) | r_T=200 MSE: 4.78e0 -> 3.54e-6 | Helped |
| Add boundary loss with uniform initialization (ablation) | Neural Adjoint (NA), Robotic arm · Robotic arm (D3) | r_T=200 MSE: 1.53e-6 -> 8.87e-7 | Helped |
| Add designed initialization without boundary loss (ablation) | Neural Adjoint (NA), Ballistics · Ballistics (D1) | r_T=200 MSE: 4.78e0 -> 1.39e0 | Hurt |
| Add designed initialization without boundary loss (ablation) | Neural Adjoint (NA), Robotic arm · Robotic arm (D3) | r_T=200 MSE: 1.53e-6 -> 5.70e-6 | Hurt |
| Add designed initialization alongside boundary loss (ablation) | Neural Adjoint (NA), Ballistics · Ballistics (D1) | r_T=200 MSE: 3.54e-6 -> 2.54e-7 | Helped |
| Add designed initialization alongside boundary loss (ablation) | Neural Adjoint (NA), Robotic arm · Robotic arm (D3) | r_T=200 MSE: 8.87e-7 -> 6.61e-7 | Helped |
| Add boundary loss (ablation) | Tandem, Ballistics and Robotic arm · Ballistics and Robotic arm | re-simulation error: at least 2.5 orders of magnitude | Helped |
| Evaluate stochastic inverse models with increasing candidate count T | Deep inverse models · Four benchmark tasks | r_T MSE | Helped |
Closest papers by meaning
- InverseBench: Benchmarking Plug-and-Play Diffusion Priors for Inverse Problems in Physical Sciences (ICLR 2025 · similarity 0.87)
- DMPlug: A Plug-in Method for Solving Inverse Problems with Diffusion Models (NeurIPS 2024 · similarity 0.86)
- A Unified Framework for Forward and Inverse Problems in Subsurface Imaging using Latent Space Translations (ICLR 2025 · similarity 0.86)
- NPN: Non-Linear Projections of the Null-Space for Imaging Inverse Problems (NeurIPS 2025 · similarity 0.86)
- Integrating Intermediate Layer Optimization and Projected Gradient Descent for Solving Inverse Problems with Diffusion Models (ICML 2025 · similarity 0.86)
- Self-diffusion for Solving Inverse Problems (NeurIPS 2025 · similarity 0.86)
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.