Yu Sun

Assistant Professor @JHUECE, Postdoc @Caltech, Ph.D. @wustlcig, Researcher in Computational imaging.

[Yu Sun @YuSunMark]

The Johns Hopkins Data Science and AI Institute @HopkinsDSAI Postdoctoral Fellowship Program is now accepting applications for the 2027–2028 academic year. Apply now! Deadline: Dec. 11, 2026. Details and apply: https://t.co/E7AFu7zI1S

[Yu Sun @YuSunMark]

🎤 I'll be giving a talk tomorrow at the IEEE SPS Webinar on diffusion models as plug-and-play priors! Welcome to join!

🔗 https://t.co/5BS8Y9ApOd

[Yu Sun @YuSunMark]

[2/2] How it works: two coupled implicit neural representations parameterize the permittivity contrast and induced current, jointly optimized under Lippmann-Schwinger physics; the score model injects a learned prior gradient via the chain rule, handling strong multiple scattering

[Yu Sun @YuSunMark]

🧵 [1/2} New preprint: ScoreField — neural electromagnetic inverse scattering with a score-based generative prior. +1.8 dB PSNR over the best baseline on real Fresnel data.
https://t.co/Nz49Re7g05

[Yu Sun @YuSunMark]

[2/2] The theory is comprehensive: convergence analysis covering single-block and multi-block implementations plus annealing schedules — under transparent assumptions, even for non-log-concave posteriors and imperfect learned score models.

[Yu Sun @YuSunMark]

[1/2] New preprint 🎉 PiX-MC: time-parallel posterior sampling for Bayesian imaging. Proximal Langevin + Picard iteration brings multi-GPU parallelism with convergence guarantees. 50x faster on sparse-view CT (8 GPUs) than vanilla Langevin (1 GPU) https://t.co/vsMD8YeLhm

[Yu Sun @YuSunMark]

[3/3] Reconstructions stay sharp under severe undersampling. We also characterize the bias from likelihood guidance and propose a ground-truth-free metric for posterior sampling quality.

[Yu Sun @YuSunMark]

🧵 [1/3] New preprint: Generative Translation Priors: Bayesian Imaging with Cross-Modality Image Translation https://t.co/7opHKzkPa9
#BayesianImaging #GenerativeAI

[Yu Sun @YuSunMark]

[2/3] The core idea is translation: a diffusion model that translates between modalities becomes the prior. Translation knowledge from one modality sharpens reconstruction in another.

[Yu Sun @YuSunMark]

@ShiqianMa Welcome! Look forward to chatting!

[Yu Sun @YuSunMark]

🎓 Congratulations to my four graduating MS students — Haoyue Guan, Yuan Gao, Xuzhe Zha, and Yetao He! Proud to see them head off to outstanding PhD programs: Haoyue to Emory, Yuan to UNC, Xuzhe to JHU, and Yetao to UCLA. Wishing them all the best in their journeys ahead! 🌟

[Yu Sun @YuSunMark]

@Mingxuan_Cai_ Thanks

[Yu Sun @YuSunMark]

Happy to announce that I received the NSF CAREER Award 🤓
https://t.co/kQUAxlCKyE

[Yu Sun @YuSunMark]

@JiYiLight Thank you, Ji!

[Yu Sun @YuSunMark]

@David_Ouyang Thanks David!

[Yu Sun @YuSunMark]

Paid summer intern opportunities for undergraduate at JHU DSAI!
https://t.co/0xgjf7zs7l

[Yu Sun @YuSunMark]

If you work on computational imaging / synthetic apertures: submit to CISA at Albany, 26-29 May 2026!
✅ 2-page abstract • ⏰ Feb 27 deadline (may extend)
Keynotes: https://t.co/SNUWqpd9Oe
Details: https://t.co/UemlqlZUGI
RT appreciated 🙏

[Yu Sun @YuSunMark]

PSI3D is in at IEEE Asilomar 2025: Plug-and-Play 3D Stochastic Inference with a Slice-wise Latent Diffusion Prior.
Scalable 3D reconstruction by pairing your forward model with a 2D diffusion prior—it turns out to suffice!
Paper: https://t.co/8Nfj8YS9qt

[Yu Sun @YuSunMark]

Got accepted to @ieeeICASSP 2026! Spain or Rio or Both? 😆🤔

[Yu Sun @YuSunMark]

Excited to announce PRISM — a probabilistic, robust inverse solver using measurement‑conditioned diffusion priors for blind inverse problems. Link: https://t.co/o91Nb2IguC. Big thanks to my students & collaborators for the great work! #ComputationalImaging #GenerativeAI

[Yu Sun @YuSunMark]

Got accepted to ICLR 2026! Congrats everyone 🏅🪇

[Yu Sun @YuSunMark]

New paper: “Provably Accelerated Imaging with Restarted Inertia & Score‑based Image Priors” — we prove FASTER convergence under general score priors! https://t.co/VT6LbhWMAP. Big thanks to my postdoc Deliang and collaborators Marien & Julien! #ComputationalImaging