Ulugbek Kamilov: "Computational Imaging: Reconciling Models and Learning"
Deep Learning and Medical Applications 2020
"Computational Imaging: Reconciling Models and Learning"
Ulugbek Kamilov, Washington University in St. Louis
Abstract: There is a growing need in biological, medical, and materials imaging research to recover information lost during data acquisition. There are currently two distinct viewpoints on addressing such information loss: model-based and learning-based. Model-based methods leverage analytical signal properties (such as wavelet sparsity) and often come with theoretical guarantees and insights. Learning-based methods leverage flexible representations (such as convolutional neural nets) for best empirical performance through training on big datasets. The goal of this talk is to introduce a framework that reconciles both viewpoints by providing the "deep learning" counterpart of the classical optimization theory. This is achieved by specifying “denoising deep neural nets” as a mechanism to infuse learned priors into recovery problems, while maintaining a clear separation between the prior and physics-based acquisition models. Our methodology can fully leverage the flexibility offered by deep learning by designing learned denoisers to be used within our new family of fast iterative algorithms. Yet, our results indicate that the such algorithms can achieve state-of-the-art performance in different computational imaging tasks, while also being amenable to rigorous theoretical analysis. We will focus on the application of the methodology to the problem to various biomedical imaging modalities, such as magnetic resonance imaging and optical tomographic microscopy.
This talk will be based on the following references:
Y. Sun, B. Wohlberg, and U. S. Kamilov, “An Online Plug-and-Play Algorithm for Regularized Image Reconstruction,” IEEE Trans. Comput. Imag., 2019. https://ieeexplore.ieee.org/document/8616843
Y. Sun, J. Liu, and U. S. Kamilov, “Block Coordinate Regularization by Denoising,” Proc. Ann. Conf. Neural Information Processing Systems (NeurIPS 2019) (Vancouver, Canada, Dec 8-14). http://papers.nips.cc/paper/8330-block-coordinate-regularization-by-denoising.pdf
Z. Wu, Y. Sun, J. Liu, and U. S. Kamilov, “Online Regularization by Denoising with Applications to Phase Retrieval,” Proc. IEEE Int. Conf. Comp. Vis. Workshops (ICCVW 2019) (Seoul, South Korea, Oct 27 – Nov 2). http://openaccess.thecvf.com/content_ICCVW_2019/html/LCI/Wu_Online_Regularization_by_Denoising_with_Applications_to_Phase_Retrieval_ICCVW_2019_paper.html
Institute for Pure and Applied Mathematics, UCLA
January 28, 2020
For more information: http://www.ipam.ucla.edu/dlm2020
Видео Ulugbek Kamilov: "Computational Imaging: Reconciling Models and Learning" канала Institute for Pure & Applied Mathematics (IPAM)
"Computational Imaging: Reconciling Models and Learning"
Ulugbek Kamilov, Washington University in St. Louis
Abstract: There is a growing need in biological, medical, and materials imaging research to recover information lost during data acquisition. There are currently two distinct viewpoints on addressing such information loss: model-based and learning-based. Model-based methods leverage analytical signal properties (such as wavelet sparsity) and often come with theoretical guarantees and insights. Learning-based methods leverage flexible representations (such as convolutional neural nets) for best empirical performance through training on big datasets. The goal of this talk is to introduce a framework that reconciles both viewpoints by providing the "deep learning" counterpart of the classical optimization theory. This is achieved by specifying “denoising deep neural nets” as a mechanism to infuse learned priors into recovery problems, while maintaining a clear separation between the prior and physics-based acquisition models. Our methodology can fully leverage the flexibility offered by deep learning by designing learned denoisers to be used within our new family of fast iterative algorithms. Yet, our results indicate that the such algorithms can achieve state-of-the-art performance in different computational imaging tasks, while also being amenable to rigorous theoretical analysis. We will focus on the application of the methodology to the problem to various biomedical imaging modalities, such as magnetic resonance imaging and optical tomographic microscopy.
This talk will be based on the following references:
Y. Sun, B. Wohlberg, and U. S. Kamilov, “An Online Plug-and-Play Algorithm for Regularized Image Reconstruction,” IEEE Trans. Comput. Imag., 2019. https://ieeexplore.ieee.org/document/8616843
Y. Sun, J. Liu, and U. S. Kamilov, “Block Coordinate Regularization by Denoising,” Proc. Ann. Conf. Neural Information Processing Systems (NeurIPS 2019) (Vancouver, Canada, Dec 8-14). http://papers.nips.cc/paper/8330-block-coordinate-regularization-by-denoising.pdf
Z. Wu, Y. Sun, J. Liu, and U. S. Kamilov, “Online Regularization by Denoising with Applications to Phase Retrieval,” Proc. IEEE Int. Conf. Comp. Vis. Workshops (ICCVW 2019) (Seoul, South Korea, Oct 27 – Nov 2). http://openaccess.thecvf.com/content_ICCVW_2019/html/LCI/Wu_Online_Regularization_by_Denoising_with_Applications_to_Phase_Retrieval_ICCVW_2019_paper.html
Institute for Pure and Applied Mathematics, UCLA
January 28, 2020
For more information: http://www.ipam.ucla.edu/dlm2020
Видео Ulugbek Kamilov: "Computational Imaging: Reconciling Models and Learning" канала Institute for Pure & Applied Mathematics (IPAM)
Показать
Комментарии отсутствуют
Информация о видео
3 апреля 2020 г. 0:05:23
00:50:53
Другие видео канала
Jaafar El-Awady - dislocation in high thermomechanical condition in Additive Manufacturing of AlloysVikram Gavini - Fast, Accurate and Large-scale Ab-initio Calculations for Materials ModelingBistra Dilkina - Machine Learning for MIP Solving - IPAM at UCLAAmit Acharya - Slow time-scale behavior of fast microscopic dynamics - IPAM at UCLAEran Rabani - Stochastic Density Functional Theory - IPAM at UCLADeanna Needell - Using Algebraic Factorizations for Interpretable Learning - IPAM at UCLAXavier Bresson - Learning to Untangle Genome Assembly Graphs - IPAM at UCLAJack Gilbert: "Microbiome of the Built Environment"John Harrison - Formalization and Automated Reasoning: A Personal and Historical PerspectiveRaymond Clay - Machine Learning in Equation of State and Transport Modeling at Extreme ConditionsJan Hermann - Neural-network wave functions for quantum chemistry - IPAM at UCLARose Yu - Incorporating Symmetry for Learning Spatiotemporal Dynamics - IPAM at UCLAYongsoo Yang - Neural network-assisted atomic electron tomography - IPAM at UCLAAlbert Fannjiang - From Tomographic Phase Retrieval to Projection Tomography - IPAM at UCLAThomas Swinburne - Learning uncertainty-aware models of defect kinetics at scale - IPAM at UCLAKevin Kelly - Machine Learning Enhanced Compressive Hyperspectral Imaging - IPAM at UCLADemetri Psaltis - Machine Learning for 3D Optical Imaging - IPAM at UCLAPaola Gori-Giorgi - Large-coupling strength expansion in DFT and Hartree-Fock adiabatic connectionsBohua Zhan - Verifying symbolic computation in the HolPy theorem prover - IPAM at UCLAXiantao Li - A stochastic algorithm for self-consistent calculations in DFT - IPAM at UCLAPascal Van Hentenryck - Fusing Machine Learning and Optimization - IPAM at UCLA