AI Paper Atlas

Research map / Image generation / Image restoration

Uncertainty Quantification via Neural Posterior Principal Components

Elias Nehme, Omer Yair, Tomer Michaeli

NeurIPS 2023 · Image generation › Image restoration · approach: Real-world restoration with unaligned data · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
use NPPC instead of a point mass at the conditional meanposterior uncertainty model · 2D Gaussian mixture denoisingWasserstein-2 distance: 4.05 -> 0.04Improved
use NPPC instead of point mass baseline with K=3 componentsposterior uncertainty model · 100-dim Gaussian mixture denoisingWasserstein-2 distance: 180.6 -> 30.3Improved
use NPPC instead of point mass baseline with K=12 componentsposterior uncertainty model · 100-dim Gaussian mixture denoisingWasserstein-2 distance: 325.7 -> 28.8Improved
use NPPC instead of DDRM sample PCAPC-based uncertainty · CelebA-HQ/FFHQ 8x noisy SRresidual error magnitude: 16.00 -> 15.11Improved
use NPPC instead of DDNM sample PCAPC-based uncertainty · CelebA-HQ/FFHQ 4x noiseless SRresidual error magnitude: 8.38 -> 8.24Improved
use NPPC instead of RePaint sample PCAPC-based uncertainty · CelebA-HQ/FFHQ eyes inpaintingresidual error magnitude: 10.3 -> 11.43Hurt
use NPPC instead of MAT sample PCAPC-based uncertainty · CelebA-HQ/FFHQ mouth inpaintingresidual error magnitude: 11.74 -> 10.27Improved
use NPPC instead of MAT sampling for speedPC-based uncertainty · CelebA-HQ/FFHQ posterior-PC comparisonneural function evaluations: 100 -> 1Improved
add Stopgrad to the Gram-Schmidt projections (ablation)NPPC joint PC learning · CelebA-HQ colorizationPC consistency (cosine similarity): 0.9 across first 3 PCs, 0.83 overallImproved
train NPPC to predict the mean jointly rather than wrapping a pretrained mean estimatorNPPC · MNIST denoisingmodel size: 4xMixed
use more PCs for severely ill-posed tasks (K=10)NPPC / posterior samplers · CelebA-HQ noisy SR vs inpaintingunexplained variance fraction: 0.8 vs 0.4Mixed

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy