Solving Linear Inverse Problems Provably via Posterior Sampling with Latent Diffusion Models

Litu Rout, Negin Raoof, Giannis Daras, Constantine Caramanis, Alexandros G. Dimakis +1 more
7/2/2023
cs.LGcs.AIstat.ML

Official code: LituRout/PSLD (161 stars, last commit 2023-12-18).

Abstract

We present the first framework to solve linear inverse problems leveraging pre-trained latent diffusion models. Previously proposed algorithms (such as DPS and DDRM) only apply to pixel-space diffusion models. We theoretically analyze our algorithm showing provable sample recovery in a linear model setting. The algorithmic insight obtained from our analysis extends to more general settings often considered in practice. Experimentally, we outperform previously proposed posterior sampling algorithms in a wide variety of problems including random inpainting, block inpainting, denoising, deblurring, destriping, and super-resolution.

View on arXivView PDF

Code Implementations(1)

LituRout/PSLDOfficial100%

Posterior Sampling using Latent Diffusion - NeurIPS 2023

16119HTML, ShellJun 20, 20232 years ago

Similar Papers

Diffusion Models for Solving Inverse Problems via Posterior Sampling with Piecewise Guidance

90% match2025

Particle-Filtering-based Latent Diffusion for Inverse Problems

90% match2024

A Survey on Diffusion Models for Inverse Problems

89% match2024

Enhancing Diffusion Models for Inverse Problems with Covariance-Aware Posterior Sampling

88% match2024

Inverse Problem Sampling in Latent Space Using Sequential Monte Carlo

88% match2026

Solving Inverse Problems with Latent Diffusion Models via Hard Data Consistency

88% match1 repo872026

Cite this paper

@article{rout2023solving,
  title  = {Solving Linear Inverse Problems Provably via Posterior Sampling with Latent Diffusion Models},
  author = {Litu Rout and Negin Raoof and Giannis Daras and Constantine Caramanis and Alexandros G. Dimakis and Sanjay Shakkottai},
  year   = {2023},
  eprint = {2307.00619},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2307.00619v1},
  journal = {NEURIPS 2023 2023}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2307.00619.svg)](https://codeofpaper.com/paper/2307.00619)

Discussion