Learning to solve TV regularized problems with unrolled algorithms - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2020

Learning to solve TV regularized problems with unrolled algorithms

Abstract

Total Variation (TV) is a popular regularization strategy that promotes piece-wise constant signals by constraining the ℓ1-norm of the first order derivative of the estimated signal. The resulting optimization problem is usually solved using iterative algorithms such as proximal gradient descent, primal-dual algorithms or ADMM. However, such methods can require a very large number of iterations to converge to a suitable solution. In this paper, we accelerate such iterative algorithms by unfolding proximal gradient descent solvers in order to learn their parameters for 1D TV regularized problems. While this could be done using the synthesis formulation, we demonstrate that this leads to slower performances. The main difficulty in applying such methods in the analysis formulation lies in proposing a way to compute the derivatives through the proximal operator. As our main contribution, we develop and characterize two approaches to do so, describe their benefits and limitations, and discuss the regime where they can actually improve over iterative procedures. We validate those findings with experiments on synthetic and real data.
Fichier principal
Vignette du fichier
main.pdf (706.15 Ko) Télécharger le fichier
bold_deconv_loss_comp_0_1.pdf (207.2 Ko) Télécharger le fichier
bold_deconv_loss_comp_0_8.pdf (207.07 Ko) Télécharger le fichier
comparison_AL_A.pdf (296.04 Ko) Télécharger le fichier
comparison_prox_tv_loss.pdf (273.88 Ko) Télécharger le fichier
iter_loss_comp_0_1.pdf (210.32 Ko) Télécharger le fichier
iter_loss_comp_0_8.pdf (210.28 Ko) Télécharger le fichier
loss_comparison.pdf (213.61 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Origin : Files produced by the author(s)

Dates and versions

hal-02954181 , version 1 (30-09-2020)
hal-02954181 , version 2 (19-10-2020)

Identifiers

  • HAL Id : hal-02954181 , version 2

Cite

Hamza Cherkaoui, Jeremias Sulam, Thomas Moreau. Learning to solve TV regularized problems with unrolled algorithms. NeurIPS 2020 - 34th Conference on Neural Information Processing Systems, Dec 2020, Vancouver / Virtuel, Canada. ⟨hal-02954181v2⟩
339 View
472 Download

Share

Gmail Facebook X LinkedIn More