Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
AGEM: Solving Linear Inverse Problems via Deep Priors and Sampling
Bichuan Guo, Yuxing Han, Jiangtao Wen
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 too small proposal scale sigma_prop 0.01 instead of 1.00 (ablation) | AGEM · Signal denoising sigma_n=3.00 | RMSE (x1e-2): 0.86 -> 34.1 | Hurt |
| use sigma_prop 0.10 instead of 1.00 (ablation) | AGEM · Signal denoising sigma_n=3.00 | RMSE (x1e-2): 0.86 -> 1.20 | Hurt |
| use too large sigma_prop 10.0 or 100 (ablation) | AGEM · Signal denoising sigma_n=3.00 | convergence | Hurt |
| initial noise level 0.5I instead of 1.0I (ablation) | AGEM · Signal denoising sigma_n=3.00 | convergence | Hurt |
| vary initial noise level among 1.0I, 2.0I, 4.0I (ablation) | AGEM · Signal denoising sigma_n=3.00 | RMSE (x1e-2): 0.87 -> 0.86 | No clear effect |
| initial noise level 8.0I (ablation) | AGEM · Signal denoising sigma_n=3.00 | convergence | Hurt |
| posterior median instead of mean as estimate (ablation) | AGEM · Signal denoising sigma_n=3.00 | RMSE (x1e-2): 0.86 -> 0.87 | No clear effect |
| use last sample instead of posterior mean (ablation) | AGEM · Signal denoising sigma_n=3.00 | RMSE (x1e-2): 0.86 -> 1.57 | Hurt |
| use first sample instead of posterior mean (ablation) | AGEM · Signal denoising sigma_n=3.00 | RMSE (x1e-2): 0.86 -> 1.58 | Hurt |
| symmetric Gaussian proposal instead of MALA (ablation) | AGEM · Signal denoising sigma_n=3.00 | RMSE (x1e-2): 0.86 -> 7.61 | Hurt |
| replace DMSP noise estimate by AGEM EM estimate (ablation) | AGEM vs DMSP · Image deblurring sigma_n=0.01 | PSNR (dB): 33.60 -> 34.79 | Improved |
| AGEM vs DMSP on devignetting (ablation) | AGEM · CelebA devignetting sigma_n=0.015 | PSNR (dB): 35.78 -> 36.34 | Improved |
Closest papers by meaning
- Stochastic Deep Restoration Priors for Imaging Inverse Problems (ICML 2025 · similarity 0.89)
- Exact Posterior Score Estimation for Solving Linear Inverse Problems (NeurIPS 2026 · similarity 0.88)
- Divide-and-Conquer Posterior Sampling for Denoising Diffusion priors (NeurIPS 2024 · similarity 0.88)
- Adapt and Diffuse: Sample-adaptive Reconstruction via Latent Diffusion Models (ICML 2024 · similarity 0.87)
- Unleashing the Denoising Capability of Diffusion Prior for Solving Inverse Problems (NeurIPS 2024 · similarity 0.87)
- A Restoration Network as an Implicit Prior (ICLR 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.
