47,778 Hits in 5.7 sec

Integrating Deep Neural Networks with Full-waveform Inversion: Reparametrization, Regularization, and Uncertainty Quantification [article]

Weiqiang Zhu, Kailai Xu, Eric Darve, Biondo Biondi, Gregory C. Beroza
2021 arXiv   pre-print
We propose a neural-network-based full waveform inversion method (NNFWI) that integrates deep neural networks with FWI by representing the velocity model with a generative neural network.  ...  The velocity model generated by neural networks is input to the same partial differential equation (PDE) solvers used in conventional FWI.  ...  layer, and a group of optimization algorithms in deep learning.  ... 
arXiv:2012.11149v3 fatcat:iptj2ibv7ndaxfwa3nzsp2dq7u

Regularized Training of Intermediate Layers for Generative Models for Inverse Problems [article]

Sean Gunn, Jorio Cocola, Paul Hand
2022 arXiv   pre-print
In our work, we introduce a principle that if a generative model is intended for inversion using an algorithm based on optimization of intermediate layers, it should be trained in a way that regularizes  ...  We instantiate this principle for two notable recent inversion algorithms: Intermediate Layer Optimization and the Multi-Code GAN prior.  ...  Deep generative models have demonstrated remarkable performance when used as priors for solving inverse problems [6, 4, 12, 24, 23, 25, 26] .  ... 
arXiv:2203.04382v1 fatcat:64qxbstquzdffkigrz3modrw3i

Invert to Learn to Invert [article]

Patrick Putzky, Max Welling
2019 arXiv   pre-print
Iterative learning to infer approaches have become popular solvers for inverse problems.  ...  As a result, the proposed approach allows us to train models with 400 layers on 3D volumes in an MRI image reconstruction task.  ...  Patrick Putzky is supported by the Netherlands Organisation for Scientific Research (NWO) and the Netherlands Institute for Radio Astronomy (ASTRON) through the big bang, big data grant.  ... 
arXiv:1911.10914v1 fatcat:tyxhrfx2xfhbfo3pjhec2nhrri

Deep Neural Networks are Surprisingly Reversible: A Baseline for Zero-Shot Inversion [article]

Xin Dong, Hongxu Yin, Jose M. Alvarez, Jan Kautz, Pavlo Molchanov
2021 arXiv   pre-print
Most existing work relies on priors or data-intensive optimization to invert a model, yet struggles to scale to deep architectures and complex datasets.  ...  20 layers.  ...  Each inversion layer is optimized via Adam [76] for 6K iterations. We use 5K synthetic images as detailed in Sec. 3.4 for the optimization. See appendix for additional experimental details.  ... 
arXiv:2107.06304v1 fatcat:ohamubvcjffxdlbe7sbioxpu2y

Deep Inverse Optimization [article]

Yingcong Tan, Andrew Delong, Daria Terekhov
2018 arXiv   pre-print
Our method, called deep inverse optimization, is to unroll an iterative optimization process and then use backpropagation to learn parameters that generate the observations.  ...  We cast inverse optimization as a form of deep learning.  ...  Our method, called deep inverse optimization, is to unroll an iterative optimization process and then use backpropagation to learn model parameters that generate the observations/targets.  ... 
arXiv:1812.00804v1 fatcat:5svkudvgyfeefcuzcmvhvtsqpu

Image Processing Using Multi-Code GAN Prior [article]

Jinjin Gu, Yujun Shen, Bolei Zhou
2020 arXiv   pre-print
In particular, we employ multiple latent codes to generate multiple feature maps at some intermediate layer of the generator, then compose them with adaptive channel importance to recover the input image  ...  We further analyze the properties of the layer-wise representation learned by GAN models and shed light on what knowledge each layer is capable of representing.  ...  We compare our inversion method with optimizing the intermediate feature maps [3] . We also compare with DIP [39] , which uses a discriminative model as prior, and Zhang et al.  ... 
arXiv:1912.07116v2 fatcat:cxb6ezakbre6hhktd6hmctpk2u

A hierarchical approach to deep learning and its application to tomographic reconstruction [article]

Lin Fu, Bruno De Man
2019 arXiv   pre-print
Here we present a novel framework to solve such problems with deep learning by casting the original problem as a continuum of intermediate representations between the input and output data.  ...  More broadly, hierarchical DL opens the door to a new class of solvers for general inverse problems, which could potentially lead to improved signal-to-noise ratio, spatial resolution and computational  ...  More broadly, it is conceivable to generalize the same methodology for solving any large-scale inverse problems and matrix decompositions.  ... 
arXiv:1912.07743v1 fatcat:xkemqw4eozfldgnp3t47qtps3q

Image Processing Using Multi-Code GAN Prior

Jinjin Gu, Yujun Shen, Bolei Zhou
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
the reconstruction from fixed PGGAN [23] models.  ...  Reconstruction (b) Image Colorization (c) Image Super-Resolution (d) Image Denoising (e) Image Inpainting (f) Semantic Manipulation Figure 1: Multi-code GAN prior facilitates many image processing applications using  ...  We compare our inversion method with optimizing the intermediate feature maps [3] . We also compare with DIP [39] , which uses a discriminative model as prior, and Zhang et al.  ... 
doi:10.1109/cvpr42600.2020.00308 dblp:conf/cvpr/GuSZ20 fatcat:vlpany75lrfifkfd46jrq6eqyq

Tutorial on the Use of Deep Learning in Diffuse Optical Tomography

Ganesh M. Balasubramaniam, Ben Wiesel, Netanel Biton, Rajnish Kumar, Judy Kupferman, Shlomi Arnon
2022 Electronics  
Diffuse optical tomography using deep learning is an emerging technology that has found impressive medical diagnostic applications.  ...  This article discusses the current state-of-the-art diffuse optical tomography systems and comprehensively reviews the deep learning algorithms used in image reconstruction.  ...  Deep learning techniques used for inverse problem solving can be broadly classified into two distinct types.  ... 
doi:10.3390/electronics11030305 fatcat:bui7xkzajvaoblrd4ttmw3odua

A novel guided deep learning algorithm to design low-cost SPP films [article]

Yingshi Chen, Jinfeng Zhu
2020 arXiv   pre-print
To achieve this goal, we use low cost sample replacement algorithm in training process. The deep CNN would gradually learn better model from samples with lower cost.  ...  The design of surface plasmon polaritons (SPP) films is an ill-posed inverse problem. There are many-to-one correspondence between the structures and user needs.  ...  We would use its' full differentiability to training and learning just like deep CNN. And its' tree model has more generality than the classical deep CNN.  ... 
arXiv:1912.03452v2 fatcat:ahdypmefxjaytbkhm75msxiueu

Face Synthesis with Landmark Points from Generative Adversarial Networks and Inverse Latent Space Mapping [article]

Shabab Bazrafkan, Hossein Javidnia, Peter Corcoran
2018 arXiv   pre-print
The inverse of the generator is implemented using an Adam optimizer to generate the latent vector corresponding to each facial image, and a lightweight deep neural network is trained to map latent Z-space  ...  This paper presents a framework for augmenting a dataset in a latent Z-space and applied to the regression problem of generating a corresponding set of landmarks from a 2D facial dataset.  ...  Project ID: 13/SPP/I2868 on "Next Generation  ... 
arXiv:1802.00390v1 fatcat:7nig425lhfcuni6coyv453qiku

Cascade Deep Networks for Sparse Linear Inverse Problems

Huan Zhang, Hong Shi, Wenwu Wang
2018 2018 24th International Conference on Pattern Recognition (ICPR)  
Sparse deep networks have been widely used in many linear inverse problems, such as image super-resolution and signal recovery.  ...  However, when the linear inverse problems involve several linear transformations or the ratio of input dimension to output dimension is large, the performance of a single sparse deep network is poor.  ...  However, the above approaches have two major limitations: 1) for the linear inverse problem, the above methods do not make full use of intermediate information when the original signal is transformed several  ... 
doi:10.1109/icpr.2018.8545232 dblp:conf/icpr/ZhangSW18 fatcat:oq7b5sw7xnfbjainldzt6njfp4

Approximate Inverse Reinforcement Learning from Vision-based Imitation Learning [article]

Keuntaek Lee, Bogdan Vlahov, Jason Gibson, James M. Rehg, Evangelos A. Theodorou
2021 arXiv   pre-print
We use Imitation Learning as a means to do Inverse Reinforcement Learning in order to create an approximate cost function generator for a visual navigation challenge.  ...  The proposed methodology relies on Imitation Learning, Model Predictive Control (MPC), and an interpretation technique used in Deep Neural Networks.  ...  CONCLUSION We introduced an Approximate Inverse Reinforcement Learning framework using deep CNNs.  ... 
arXiv:2004.08051v3 fatcat:qmqa2uoirbhxdfrra3wrva5w6i

Image Reconstruction by Splitting Deep Learning Regularization from Iterative Inversion [chapter]

Jiulong Liu, Tao Kuang, Xiaoqun Zhang
2018 Lecture Notes in Computer Science  
In this work, we propose a general and easy-to-use reconstruction method based on deep learning techniques.  ...  In order to address the intractable inversion of general inverse problems, we propose to train a network to refine intermediate images from classical reconstruction procedure to the ground truth, i.e.  ...  In [8, 6, 9] , analytic solutions are obtained for the inversion layers and a proximal operator is learned for the denoising/anti-artifact layers.  ... 
doi:10.1007/978-3-030-00928-1_26 fatcat:qefyg7boqjc4lkrsg62kbvsf4q

Deep CNN Denoiser and Multi-layer Neighbor Component Embedding for Face Hallucination

Junjun Jiang, Yi Yu, Jinhui Hu, Suhua Tang, Jiayi Ma
2018 Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence  
To solve this problem, in this paper we propose a general face hallucination method that can integrate model-based optimization and discriminative inference.  ...  In particular, to exploit the model based prior, the Deep Convolutional Neural Networks (CNN) denoiser prior is plugged into the super-resolution optimization model with the aid of image-adaptive Laplacian  ...  ., 2017] take advantage of Deep CNN discriminative learning and incorporated it to the model-based optimization methods to tackle with the inverse problems.  ... 
doi:10.24963/ijcai.2018/107 dblp:conf/ijcai/Jiang0HT018 fatcat:nawjw4pfufffpfii2e73rousqm
« Previous Showing results 1 — 15 out of 47,778 results