NETT: Solving Inverse Problems with Deep Neural Networks

@article{Li2018NETTSI,
  title={NETT: Solving Inverse Problems with Deep Neural Networks},
  author={Housen Li and Johannes Schwab and Stephan Antholzer and Markus Haltmeier},
  journal={ArXiv},
  year={2018},
  volume={abs/1803.00092}
}
Recovering a function or high-dimensional parameter vector from indirect measurements is a central task in various scientific areas. Several methods for solving such inverse problems are well developed and well understood. Recently, novel algorithms using deep learning and neural networks for inverse problems appeared. While still in their infancy, these techniques show astonishing performance for applications like low-dose CT or various sparse data problems. However, there are few theoretical… Expand
Deep null space learning for inverse problems: convergence analysis and rates
Recently, deep learning based methods appeared as a new paradigm for solving inverse problems. These methods empirically show excellent performance but lack of theoretical justification; inExpand
Regularization of Inverse Problems by Neural Networks
TLDR
It is shown that neural networks can be used for the data-driven solution of inverse problems and existing deep learning methods for inverse problems are viewed from the perspective of regularization theory, the mathematical foundation of stable solution methodsfor inverse problems. Expand
Data-consistent neural networks for solving nonlinear inverse problems
TLDR
This paper proposes data-consistent neural networks that are combined with classical regularization methods that provide a stable solution of inverse problems that exploits both the known nonlinear forward model as well as the desired solution manifold from data. Expand
Learning regularization parameters of inverse problems via deep neural networks
TLDR
A new approach that uses deep neural networks (DNN) to obtain regularization parameters for solving inverse problems and avoids the need for expensive and challenging bilevel optimization methods as utilized in other existing training approaches. Expand
Discretization of Learned NETT Regularization for Solving Inverse Problems
TLDR
The asymptotic convergence of the discretized NETT approach for decreasing noise levels and discretization errors is shown and the derive convergence rates and present numerical results for a limited data problem in photoacoustic tomography. Expand
Deep synthesis regularization of inverse problems
TLDR
Deep synthesis regularization (DESYRE) is introduced using neural networks as nonlinear synthesis operator bridging the gap between these two worlds and allows to exploit the deep learning benefits of being well adjustable to available training data. Expand
Deep learning architectures for nonlinear operator functions and nonlinear inverse problems
TLDR
A new parametric family is introduced that resembles a standard neural network architecture, but where the input data acts multiplicatively on vectors, and it is observed that the sparsity of the weight matrices improves the generalization estimates. Expand
Neural Networks-based Regularization of Large-Scale Inverse Problems in Medical Imaging
TLDR
A generalized Deep Learning-based approach to solve ill-posed large-scale inverse problems occurring in medical imaging by decoupling the regularization of the solution from ensuring consistency with the measured data and which outperformed all the reported methods with respect to all chosen quantitative measures. Expand
Solving Inverse Problems With Deep Neural Networks - Robustness Included?
TLDR
An extensive study of the robustness of deep-learning-based algorithms for solving underdetermined inverse problems covers compressed sensing with Gaussian measurements as well as image recovery from Fourier and Radon measurements, including a real-world scenario for magnetic resonance imaging. Expand
Deep neural networks learning to solve nonlinear inverse problems for the wave equation
We present the construction of a deep neural network for solving the nonlinear inverse problem of determining an unknown piecewise smooth wave speed from boundary measurements, where wave propagationExpand
...
1
2
3
4
5
...

References

SHOWING 1-10 OF 95 REFERENCES
Solving ill-posed inverse problems using iterative deep neural networks
TLDR
The method builds on ideas from classical regularization theory and recent advances in deep learning to perform learning while making use of prior information about the inverse problem encoded in the forward operator, noise model and a regularizing functional to results in a gradient-like iterative scheme. Expand
One Network to Solve Them All — Solving Linear Inverse Problems Using Deep Projection Models
TLDR
This work proposes a general framework to train a single deep neural network that solves arbitrary linear inverse problems and demonstrates superior performance over traditional methods using wavelet sparsity prior while achieving performance comparable to specially-trained networks on tasks including compressive sensing and pixel-wise inpainting. Expand
MoDL: Model-Based Deep Learning Architecture for Inverse Problems
TLDR
This work introduces a model-based image reconstruction framework with a convolution neural network (CNN)-based regularization prior, and proposes to enforce data-consistency by using numerical optimization blocks, such as conjugate gradients algorithm within the network. Expand
Adversarial Regularizers in Inverse Problems
TLDR
This work proposes a new framework for applying data-driven approaches to inverse problems, using a neural network as a regularization functional, that can be applied even if only unsupervised training data is available. Expand
Deep Convolutional Neural Network for Inverse Problems in Imaging
TLDR
The proposed network outperforms total variation-regularized iterative reconstruction for the more realistic phantoms and requires less than a second to reconstruct a <inline-formula> <tex-math notation="LaTeX">$512\times 512$ </tex- math></inline- formula> image on the GPU. Expand
Deep learning for photoacoustic tomography from sparse data
TLDR
A direct and highly efficient reconstruction algorithm based on deep learning is developed for the sparse data problem in photoacoustic tomography (PAT), which reconstructs images with a quality comparable to state of the art iterative approaches for PAT from sparse data. Expand
Sparse synthesis regularization with deep neural networks
TLDR
An encoder-decoder network is trained by including an ℓ1-penalty and it is demonstrated that the trained decoder network allows sparse signal reconstruction using thresholded encoded coefficients without losing much quality of the original image. Expand
Input Convex Neural Networks
This paper presents the input convex neural network architecture. These are scalar-valued (potentially deep) neural networks with constraints on the network parameters such that the output of theExpand
NETT regularization for compressed sensing photoacoustic tomography
TLDR
The deep learning method of [H. Li, J. Schwab, S. Antholzer, and M. Haltmeier] is applied for the first time to the CS-PAT problem, and a network architecture and training strategy for the NETT is proposed that is expected to be useful for other inverse problems as well. Expand
A U-Nets Cascade for Sparse View Computed Tomography
TLDR
A new convolutional neural network architecture for image reconstruction in sparse view computed tomography using a cascade of U-nets and data consistency layers yields superior visual results and better preserves the overall image structure as well as fine diagnostic details, e.g. the coronary arteries. Expand
...
1
2
3
4
5
...