Learning-based methods, and in particular deep neural networks, have emerged as highly successful and universal tools for image and signal recovery and restoration. They achieve state-of-the-art results on tasks ranging from image denoising, image compression, and image reconstruction from few and noisy measurements. They are starting to be used in important imaging technologies, for example in GEs newest computational tomography scanners and in the newest generation of the iPhone. They are also used in inverse problems beyond imaging, including for solving inverse problems arrising in communications, signal processing, and even on non-Euclidean data suchs as graphs.
The field has a range of theoretical and practical questions that remain unanswered, including questions about guarantees, robustness, architectural design, the role of learning, domain specific applications, and more. This virtual workshop aims at bringing together theoreticians and practitioners in order to chart out recent advances and discuss new directions in deep learning-based approaches for solving inverse problems in the imaging sciences and beyond.
How to attend the workshop
The workshop is livestreamed at the NeurIPS workshop webpage. To access this website you need to register for the conference at https://neurips.cc/ and be logged in.
The livestream will play the talks, and each of the talks is followed by a live discussion. You can attend the live discussion either through the live stream or you can join the zoom session. Questions should be asked either through rocketchat or in the chat function in the zoom session, and a co-organizer will moderate.
During the designated discussion sessions you can talk to the respective speakers in gather.town.
The poster session is also on gather.town. The gather.town link is accessible only when logged in at neurips.cc and registered for the conference.
The workshop schedule is aligned with 7:55 AM to 4 PM PT; please see this converter for conversion to your specific time zone.
|8:00||Alex Dimakis (UT Austin) Talk|
|8:30||Muyinatu Bell (John Hopkins University) Talk|
|9:00||Contributed talk 1: Batu Ozturkler, Arda Sahiner, Tolga Ergen, Arjun D Desai, John M. Pauly, Shreyas Vasanawala, Morteza Mardani, Mert Pilanci: Greedy Learning for Large-Scale Neural MRI Reconstruction|
|9:15||Contributed talk 2: Alireza Naderi, Yaniv Plan: Beyond Independent Measurements: General Compressed Sensing with GNN Application|
|9:30||Break and Discussion with Alex Dimakis and Muyinatu Bell at gather.town|
|10:00||Contributed talk 3: Eric Markley, Fanglin Linda Liu, Michael Kellman, Nick Antipa, Laura Waller: Physics-Based Learned Diffuser for Single-shot 3D Imaging|
|10:15||Kerstin Hammernik (TUM) Talk|
|10:45||Carola Schönlieb (University of Cambridge) Talk|
|11:15||Break and discussion with Kerstin Hammernik and Carola Schönlieb at gather.town|
|12:45||Poster session at gather.town|
|2:00||Gordon Wetzstein (Stanford) Talk|
|2:30||Stan Osher (UCLA) Talk|
|3:00||Jonas Adler (DeepMind) Talk|
|3:30||Discussion with Gordon Wetzstein, Stan Osher, and Jonas Adler at gather.town|
|4:00||End of official program|
John Hopkins University
Technical University of Munich
All accepted papers are available on openreview:
A Closer Look at Reference Learning for Fourier Phase Retrieval
Tobias Uelwer, Nick Rucks, Stefan Harmeling
Bayesian Inference in Physics-Based Nonlinear Flame Models
Maximilian L. Croci, Ushnish Sengupta, Matthew P Juniper
Beyond Independent Measurements: General Compressed Sensing with GNN Application
Alireza Naderi, Yaniv Plan
DARTS for Inverse Problems: a Study on Stability
Jonas Geiping, Jovita Lukasik, Margret Keuper, Michael Moeller
Deep subspace learning for efficient reconstruction of spatiotemporal imaging data
Christopher Michael Sandino, Frank Ong, Siddharth Srinivasan Iyer, Adam Bush, Shreyas Vasanawala
Efficient posterior inference & generalization in physics-based Bayesian inference with conditional GANs
Deep Ray, Dhruv V Patel, Harisankar Ramaswamy, Assad Oberai
Functional Response Conditional Variational Auto-Encoders for Inverse Design of Metamaterials
Che Wang, Yuhao Fu, Ke Deng, Chunlin Ji
Greedy Learning for Large-Scale Neural MRI Reconstruction
Batu Ozturkler, Arda Sahiner, Tolga Ergen, Arjun D Desai, John M. Pauly, Shreyas Vasanawala, Morteza Mardani, Mert Pilanci
Invertible Learned Primal-Dual
Jevgenija Rudzusika, Buda Bajic, Ozan Öktem, Carola-Bibiane Schönlieb, Christian Etmann
Learning convex regularizers satisfying the variational source condition for inverse problems
Subhadip Mukherjee, Carola-Bibiane Schönlieb, Martin Burger
Learning Lipschitz-Controlled Activation Functions in Neural Networks for Plug-and-Play Image Reconstruction Methods
Pakshal Bohra, Dimitris Perdios, Alexis Goujon, Sébastien Emery, Michael Unser
Learning Structured Sparse Matrices for Signal Recovery via Unrolled Optimization
Jonathan Sauder, Martin Genzel, Peter Jung
Likelihood-Free Inference in State-Space Models with Unknown Dynamics
Alexander Aushev, Thong Anh Tran, Henri Pesonen, Andrew Howes, Samuel Kaski
Matching Plug-and-Play Algorithms to the Denoiser
Saurav K Shastri, Rizwan Ahmad, Christopher Metzler, Philip Schniter
Multi-Task Accelerated MR Reconstruction Schemes for Jointly Training Multiple Contrasts
Victoria Liu, Kanghyun Ryu, Cagan Alkan, John M. Pauly, Shreyas Vasanawala
Near-Exact Recovery for Sparse-View CT via Data-Driven Methods
Martin Genzel, Ingo Gühring, Jan Macdonald, Maximilian März
NeRP: Implicit Neural Representation Learning with Prior Embedding for Sparsely Sampled Image Reconstruction
Liyue Shen, John M. Pauly, Lei Xing
PANOM: Automatic Hyper-parameter Tuning for Inverse Problems
Tianci Liu, Quan Zhang, Qi Lei
Photoacoustic imaging with conditional priors from normalizing flows
Rafael Orozco, Ali Siahkoohi, Gabrio Rizzuti, Tristan van Leeuwen, Felix Johan Herrmann
Photon-Limited Deblurring using Algorithm Unrolling
Yash Sanghvi, Abhiram Gnanasambandam, Stanley Chan
Physics-Based Learned Diffuser for Single-shot 3D Imaging
Eric Markley, Fanglin Linda Liu, Michael Kellman, Nick Antipa, Laura Waller
PLUGIn-CS: A simple algorithm for compressive sensing with generative prior
Babhru Joshi, Xiaowei Li, Yaniv Plan, Ozgur Yilmaz
ReFIn: A Refinement Approach for Video Frame Interpolation
Saikat Dutta, Anurag Mittal
Robust Compressed Sensing MR Imaging with Deep Generative Priors
Ajil Jalal, Marius Arvinte, Giannis Daras, Eric Price, Alex Dimakis, Jonathan Tamir
Solving Inverse Problems in Medical Imaging with Score-Based Generative Models
Yang Song, Liyue Shen, Lei Xing, Stefano Ermon
Sparse deep computer-generated holography for optical microscopy
Alex Liu, Yi Xue, Laura Waller
SSFD: Self-Supervised Feature Distance as an MR Image Reconstruction Quality Metric
Philip M Adamson, Beliz Gunel, Jeffrey Dominic, Arjun D Desai, Daniel Spielman, Shreyas Vasanawala, John M. Pauly, Akshay Chaudhari
Zero-Shot Physics-Guided Deep Learning for Subject-Specific MRI Reconstruction
Burhaneddin Yaman, Seyed Amir Hossein Hosseini, Mehmet Akcakaya
Technical University of Munich
University of Maryland
University of Southern California
University of Chicago
Call for Papers and Submission Instructions
We invite researchers to submit anonymous papers of up to 4 pages (excluding references and appendices) which will be considered for contributed workshop papers. No specific formatting is required. Authors are encouraged to use the NeurIPS style file, but they may use any other style as long as it has standard font size (11pt) and margins (1in).
Submission at OpenReview is open now until the submission deadline on September 26, 2021.
We welcome all submissions in the intersection of inverse problems and deep learning, including but not limited to submissions on the following topics:
Deep learning based approaches can make drastic reconstruction errors and may introduce biases. How common are such reconstruction problems, can they be alleviated, and if yes, how?
Deep learning based approaches often lack the guarantees of the traditional physics based methods. What theoretical results are necessary and possible?
Coordinate-based signal representations (e.g., NeRF and SIREN) and untrained convolutional neural networks have shown that neural networks alone, without any learning, can give excellent reconstruction performance. Is it possible to achieve state-of-the art performance without or little training data?
- Submission Deadline:
September 17, 2021extended until September 26, 2021, AoE (this extension is final).
- Notification of acceptance: October 15, 2021.
- SlidesLive upload for speaker videos: November 1, 2021
- Workshop: Monday December 13, 2021.
Please email email@example.com with any questions.