Johns Hopkins' Yu Sun presents two provable diffusion-prior sampling algorithms for inverse problems
prof_kamilov · x · 2026-10-22
At an IEEE SPS webinar (Oct 22), Johns Hopkins assistant professor Yu Sun will present "Provable Probabilistic Imaging using Generative Diffusion Priors."
Key points:
- Inverse problems are often ill-posed: multiple images may fit the same measurements, so solutions must both recover high-quality images and characterize reconstruction ambiguity.
- Many existing diffusion-model-based imaging approaches approximate the generative process, which can push solutions away from the true Bayesian posterior.
- He proposes two diffusion-based sampling algorithms avoiding such approximations, framed as complementary sampling extensions of classical plug-and-play priors (PnP): one based on gradient descent, the other on variable splitting.
- Both come with distributional convergence guarantees, with results on black-hole imaging showing multiple plausible solutions and uncertainty quantification in highly ill-posed settings.
More from Research
- Brain-inspired semiconductor chips aim for smarter, greener AI data centers — striketheviol · 2026-10-22
- How machine cognition research bridges to human CogSci: a Yale linguistics talk — najoungkim · 2026-10-22
- SARAL AI turns Indian public research into reels and posters for farmers — ponguru · 2026-10-22
- ETH Zurich maps economic theory papers into claim graphs with AI to weigh which errors matter — Afinetheorem · 2026-10-22
- Sheffield advertises funded PhD on reliability and bias of LLM-based evaluation — nikaletras · 2026-10-22
- New arXiv paper extracts the cruxes of LLM forecasts by contrasting 30 predictions — universeinanegg · 2026-10-22