Excited to announce my internship work @GoogleAI
has been accepted to #NeurIPS2025.
This work introduces Kernel Density Steering (KDS), a training-free method to improve the performance and stability of image-to-image diffusion models. KDS achieves this by simultaneously sampling an N-particle ensemble to derive an empirical density while applying explicit mode-seeking, which leads to better consistency and fewer artifacts.
Grateful for my incredible hosts and collaborators: @2ptmvd, @docmilanfar, @KangfuM, Mojtaba Sahraee-Ardakan and @ukmlv.
Excited to announce my work during @GoogleAI internship is now on arXiv!
Grateful for my incredible hosts and collaborators: @2ptmvd, @docmilanfar, @KangfuM, Mojtaba Sahraee-Ardakan and @ukmlv.
Please check out the paper here: [https://t.co/QL5BWQaRDw].
The @OdenInstitute Peter O'Donnell Jr. Postdoc Fellowship program is accepting applications (deadline Dec 1). Contact me if you are interested in applying and working together on 💻🧲🖼️ (computational magnetic resonance imaging)! https://t.co/nRrSLHpM0v
I am looking to hire several Ph.D. students starting Fall 2025 @JHUECE. Anyone who is passionate for computational imaging/machine learning/statistical inference is welcome to apply. MS/Postdoc positions are also available! Here is the link: https://t.co/8bs1asw7mK #PhDposition
➤ M. Renaud, J. Liu, V. de Bortoli, A. Almansa, and U. S. Kamilov, “Plug-and-Play Posterior Sampling under Mismatched Measurement and Prior Models”
Authors on X: @Jiaming__Liu@ValentinDeBort1@AndresAlmansaR@ukmlv
Link: https://t.co/cHrnpJxGT9 (3/3)
New paper "FLAIR: A Conditional Diffusion Framework with Applications to Face Video Restoration" presents a new conditional diffusion model for efficiently restoring videos using dedicated priors on faces.
⭑ Read here: https://t.co/nYiYTf5zz6
⭑ Video: https://t.co/9wjDOttDaw
Excited to share my latest work at Caltech, named ‘Provable Probabilistic Imaging using Score-Based Generative Priors’. This work is a collaboration with @ZihuiRayWu, Yifan Chen, Berthy Feng, and @klbouman.
https://t.co/yMukKOFLhQ
📢 We have an open postdoc position at ENS Lyon on multilevel unrolled & plug-and-play methods.
See details here:
https://t.co/nMi8VgwWqv
We would highly appreciate a retweet
Good news in medical generative AI!🎉🎉 We just published the preprint of the MONAI Generative Models Extension!
Check out our latest experiments with 2D and 3D data, ControlNets and 3D Cascaded Diffusion Models!
https://t.co/CRcgi17PZB
#AI#MedicalImaging#GenerativeModels
📢📢 Release of DeepInverse library 📢📢
After months of intense work, we are releasing the first stable version of DeepInverse https://t.co/o7LpxJJjdz, a PyTorch library for solving inverse problems with deep learning.
with @HuraultSamuel, @MatthieuTerris and @ddongchen
A 🧵
ICCP 2023 call for posters and demo submissions now out! Posters and demos provide an opportunity to showcase previously published or yet-to-be-published work. Due on June 15, 2023.
https://t.co/V0UKHOk8NC
CIG is seeking a postdoctoral researcher to join our team in 2023. The successful candidate will work on an exciting collaborative project at the intersection of computational imaging, neuroscience, and deep learning. Please help us to spread the word: https://t.co/q7DJ6YZJYo.
Polyak-Łojasiewicz inequality generalizes strong convexity, and a simple condition to ensure linear converge of gradient descent even for non-convex functions. You should read the very nice and clear paper of @MarkSchmidtUBC and collaborators about this!
https://t.co/dfTWCrCBXC
Our paper "Recovery Analysis for Plug-and-Play Priors using the Restricted Eigenvalue Condition" has been accepted to @NeurIPSConf 2021. We present theory that relates compressive sensing using generative models to plug-and-play priors (PnP). https://t.co/7gBvgLdAB0
New: Our overview paper on Mobile Computational Photography just appeared in the Annual Review of Vision Science. We give a brief history & describe some key technological components, including burst photography, noise reduction & super-res
🔓access here: https://t.co/NPC00WtGig
What happens to Adam if we turn off momentum and epsilon? If we set β₁=0, we get RMSprop. If we set β₁=β₂=0, we get signSGD. How well does signSGD with constant batch size converge? It doesn't. Not even with tiny stepsizes and overparameterization.
https://t.co/SobtfHj0IL