Research map / Probabilistic and scientific modeling / Variational and generative sampling
Composing Normalizing Flows for Inverse Problems
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 composed-flow SVI rather than latent-space LMC | Conditional inpainting with a RealNVP prior · MNIST | FID: 14.56 -> 5.15 | Helped |
| Use composed-flow SVI rather than latent-space LMC | Conditional inpainting with a RealNVP prior · CIFAR-10 (5-bit) | MSE: 9.31 -> 9.73 | Hurt |
| Use composed-flow SVI rather than latent-space LMC | Conditional inpainting with a RealNVP prior · CelebA-HQ (5-bit) | FID: 30.34 -> 37.24 | Hurt |
| Amortize the pre-generator over observations (ablation) | Composed-flow inference · MNIST image completion | FID: 5.15 -> 8.25 | Hurt |
| Amortize the pre-generator over observations (ablation) | Composed-flow inference · MNIST image completion | Speed (seconds/sample): 13.6 -> 0.025 | Helped |
| Amortize the pre-generator over observations (ablation) | Composed-flow inference · CelebA-HQ image completion | FID: 37.2 -> 83.1 | Hurt |
| Amortize the pre-generator over observations (ablation) | Composed-flow inference · CelebA-HQ image completion | Speed (seconds/sample): 70.5 -> 0.046 | Helped |
| Use LPIPS distance instead of Gaussian smoothing (ablation) | Composed-flow SVI · Inverse problems | Sample quality | No clear effect |
| Anneal smoothing variance from large to small positive values (ablation) | Composed-flow SVI · Inverse problems | Sample quality | No clear effect |
| Increase Gaussian smoothing variance substantially (ablation) | Composed-flow conditional sampler · Inverse problems | Observation matching | Hurt |
| Decrease Gaussian smoothing variance toward zero (ablation) | Composed-flow conditional sampler · Inverse problems | Sample quality | Hurt |
| Estimate the conditional mean from multiple samples rather than use a single sample (ablation) | Composed-flow posterior · Compressed sensing on CelebA-HQ | PSNR | Helped |
Closest papers by meaning
- Latent Refinement via Flow Matching for Training-free Linear Inverse Problem Solving (NeurIPS 2025 · similarity 0.88)
- Solving Inverse Problems with FLAIR (NeurIPS 2025 · similarity 0.88)
- Fast samplers for Inverse Problems in Iterative Refinement models (NeurIPS 2024 · similarity 0.88)
- Flow Priors for Linear Inverse Problems via Iterative Corrupted Trajectory Matching (NeurIPS 2024 · similarity 0.88)
- Exact Posterior Score Estimation for Solving Linear Inverse Problems (NeurIPS 2026 · similarity 0.87)
- CAR-Flow: Condition-Aware Reparameterization Aligns Source and Target for Better Flow Matching (NeurIPS 2025 · 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.