Research map / Probabilistic and scientific modeling / Variational and generative sampling
Posterior Matching for Arbitrary Conditioning
Ryan Strauss, Junier Oliva
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 VDVAE rather than VQ-VAE as the base model with Posterior Matching | Posterior Matching image inpainting · MNIST, Omniglot, and CelebA | PSNR | Mixed |
| Apply Posterior Matching rather than VAEAC | Tabular arbitrary conditional density estimation · Power, Gas, Hepmass, Miniboone, BSDS | NRMSE: Power: 0.880 $\pm$ 0.001 -> 0.834 $\pm$ 0.001; Gas: 0.574 $\pm$ 0.033 -> 0.330 $\pm$ 0.013; Hepmass: 0.896 $\pm$ 0.001 - | Improved |
| Apply Posterior Matching rather than VAEAC | Tabular arbitrary conditional density estimation · Power, Gas, Hepmass, Miniboone, BSDS | LL: Power: -0.042 $\pm$ 0.002 -> 0.246 $\pm$ 0.002; Gas: 2.418 $\pm$ 0.006 -> 5.964 $\pm$ 0.005; Hepmass: -10.082 $\pm$ 0.01 | Improved |
| Use VDVAE + Posterior Matching instead of ACFlow+BG | Image inpainting · MNIST and CelebA | PSNR: MNIST: 20.828 $\pm$ 0.031 -> 21.603 $\pm$ 0.022; CelebA: 25.723 $\pm$ 0.020 -> 27.190 $\pm$ 0.049 | Improved |
| Use VDVAE + Posterior Matching instead of ACFlow+BG | Image inpainting · Omniglot | PSNR: 18.838 $\pm$ 0.009 -> 18.256 $\pm$ 0.038 | Hurt |
| Replace sampling-based feature acquisition with learned lookahead posteriors | Greedy active feature acquisition · MNIST | acquisition time: 68 ms $\pm$ 917 $\mu$s -> 310 $\mu$s $\pm$ 15.3 $\mu$s; 219x | Improved |
| Use lookahead posteriors instead of sampling-based acquisition | Greedy active feature acquisition · MNIST | reconstruction RMSE | No clear effect |
Closest papers by meaning
- Variational Autoencoder with Implicit Optimal Priors (AAAI 2019 · similarity 0.87)
- Variational Learning of Fractional Posteriors (ICML 2025 · similarity 0.87)
- Arbitrary Conditional Distributions with Energy (NeurIPS 2021 · similarity 0.87)
- Beyond Vanilla Variational Autoencoders: Detecting Posterior Collapse in Conditional and Hierarchical Variational Autoencoders (ICLR 2024 · similarity 0.87)
- Undirected Graphical Models as Approximate Posteriors (ICML 2020 · similarity 0.87)
- Diffusion Model for Dense Matching (ICLR 2024 · 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.
