Designing Recurrent Neural Networks by Unfolding an L1-L1 Minimization Algorithm

Lê, Hung; Luong, Huynh Van; Deligiannis, Nikos

doi:10.1109/icip.2019.8803281

Cited by 12 publications

(12 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…By contrast [22] and [23] demonstrate that the sensing matrix training strategy and the deblocking strategy both play a key role in improving the reconstruction quality. We also noted that there are few studies on video reconstruction using unrolling methods, such as those described in [24] and [25].…”

Section: Related Workmentioning

confidence: 93%

GPX-ADMM-Net: Interpretable Deep Neural Network for Image Compressive Sensing

Lin

2021

IEEE Access

View full text Add to dashboard Cite

The building of effective neural network architectures for solving image compressive sensing (CS) problems is a challenge. Hence, it is helpful to consult the structural insight provided by traditional optimization algorithms, the results of which are interpretable. Inspired by this concept, we propose a fast, light, easily interpreted, and relatively stable learning-based method for solving image CS problems. This method is termed the Generalized ProXimal Alternating Direction Method of Multipliers Network (GPX-ADMM-Net) and can be used to solve image CS problems that conventionally are solved using optimization algorithms with intensive computations. GPX-ADMM-Net features the following characteristics: (1) It achieves a higher or comparable performance with state-of-the-art learning-based methods for the image CS problem in terms of reconstruction quality, running speed, and storage cost. (2) Because our method strictly follows the insight of optimization algorithms, it possesses a measurement rate (MR) adaptivity able to reconstruct acceptable image qualities under different MRs with only a single set of trained parameters, which is a unique property in learning-based methods. (3) Despite its dedicated design for the image CS problem, it possesses a cross-task adaptivity extended to handle other image inverse problems (e.g., superresolution and inpainting) that are treated as a special case of CS. Extensive experimental results validate the above claims.

show abstract

Section: Related Workmentioning

confidence: 93%

GPX-ADMM-Net: Interpretable Deep Neural Network for Image Compressive Sensing

Lin

2021

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Deep unfolding methods have been achieving state-of-theart performance in solving decomposition problems in terms of both accuracy and computational complexity [13], [14], [15], [16], [6], [7], [8]. The study in [13] proposed to unroll the iterations of the Iterative Shrinkage-Thresholding Algorithm (ISTA) to a feed-forward neural network-coined Learned ISTA (LISTA)-which is trained on data.…”

Section: Introductionmentioning

confidence: 99%

A Deep-Unfolded Reference-Based RPCA Network For Video Foreground-Background Separation

Luong

Joukovsky

Eldar

et al. 2021

2020 28th European Signal Processing Conference (EUSIPCO)

Self Cite

View full text Add to dashboard Cite

Deep unfolded neural networks are designed by unrolling the iterations of optimization algorithms. They can be shown to achieve faster convergence and higher accuracy than their optimization counterparts. This paper proposes a new deep-unfolding-based network design for the problem of Robust Principal Component Analysis (RPCA) with application to video foreground-background separation. Unlike existing designs, our approach focuses on modeling the temporal correlation between the sparse representations of consecutive video frames. To this end, we perform the unfolding of an iterative algorithm for solving reweighted 1-1 minimization; this unfolding leads to a different proximal operator (a.k.a. different activation function) adaptively learned per neuron. Experimentation using the moving MNIST dataset shows that the proposed network outperforms a recently proposed state-of-the-art RPCA network in the task of video foreground-background separation.

show abstract

“…Deep unfolding networks are interpretable (as opposed to traditional DNNs) and have proven to be superior to traditional optimizationbased methods and DNN models (because they integrate prior knowledge about the signal structure). Examples of such networks include LISTA [14], 1 -1 -RNN [15], and LeSITA [16]. In a recent study we presented a DeepFPC network [17] designed by unfolding the iterations of the FPC-1 algorithm.…”

Section: Introductionmentioning

confidence: 99%

A Robust Deep Unfolded Network for Sparse Signal Recovery from Noisy Binary Measurements

Yang

Xiao

Liao

et al. 2021

2020 28th European Signal Processing Conference (EUSIPCO)

Self Cite

View full text Add to dashboard Cite

We propose a novel deep neural network, coined DeepFPC-2, for solving the 1-bit compressed sensing problem. The network is designed by unfolding the iterations of the fixed-point continuation (FPC) algorithm with one-sided 2-norm (FPC-2). The DeepFPC-2 method shows higher signal reconstruction accuracy and convergence speed than the traditional FPC-2 algorithm. Furthermore, we compare its robustness to noise with the previously proposed DeepFPC network-which stemmed from unfolding the FPC-1 algorithm-for different signal to noise ratio (SNR) and sign-flipped ratio (flip ratio) scenarios. We show that the proposed network has better noise immunity than the previous DeepFPC method. This result indicates that the robustness of a deep-unfolded neural network is related with that of the algorithm it stems from.

show abstract

Designing Recurrent Neural Networks by Unfolding an L1-L1 Minimization Algorithm

Cited by 12 publications

References 20 publications

GPX-ADMM-Net: Interpretable Deep Neural Network for Image Compressive Sensing

GPX-ADMM-Net: Interpretable Deep Neural Network for Image Compressive Sensing

A Deep-Unfolded Reference-Based RPCA Network For Video Foreground-Background Separation

A Robust Deep Unfolded Network for Sparse Signal Recovery from Noisy Binary Measurements

Contact Info

Product

Resources

About