TY - GEN
T1 - Fast Inexact Bilevel Optimization for Analytical Deep Image Priors
AU - Salehi, Mohammad Sadegh
AU - Bubba, Tatiana A.
AU - Korolev, Yury
PY - 2025/5/17
Y1 - 2025/5/17
N2 - The analytical deep image prior (ADP) introduced by Dittmer et al. (2020) establishes a link between deep image priors and classical regularization theory via bilevel optimization. While this is an elegant construction, it involves expensive computations if the lower-level problem is to be solved accurately. To overcome this issue, we propose to use adaptive inexact bilevel optimization to solve ADP problems. We discuss an extension of a recent inexact bilevel method called the method of adaptive inexact descent of Salehi et al.(2024) to an infinite-dimensional setting required by the ADP framework. In our numerical experiments we demonstrate that the computational speed-up achieved by adaptive inexact bilevel optimization allows one to use ADP on larger-scale problems than in the previous literature, e.g. in deblurring of 2D color images.
AB - The analytical deep image prior (ADP) introduced by Dittmer et al. (2020) establishes a link between deep image priors and classical regularization theory via bilevel optimization. While this is an elegant construction, it involves expensive computations if the lower-level problem is to be solved accurately. To overcome this issue, we propose to use adaptive inexact bilevel optimization to solve ADP problems. We discuss an extension of a recent inexact bilevel method called the method of adaptive inexact descent of Salehi et al.(2024) to an infinite-dimensional setting required by the ADP framework. In our numerical experiments we demonstrate that the computational speed-up achieved by adaptive inexact bilevel optimization allows one to use ADP on larger-scale problems than in the previous literature, e.g. in deblurring of 2D color images.
KW - Data-driven regularization
KW - inexact bilevel optimization
KW - regularization by architecture
KW - semi-blind deblurring
UR - https://www.scopus.com/pages/publications/105006804113
UR - https://arxiv.org/abs/2502.09758
U2 - 10.1007/978-3-031-92366-1_3
DO - 10.1007/978-3-031-92366-1_3
M3 - Chapter in a published conference proceeding
AN - SCOPUS:105006804113
SN - 9783031923654
T3 - Lecture Notes in Computer Science
SP - 30
EP - 42
BT - Scale Space and Variational Methods in Computer Vision - 10th International Conference, SSVM 2025, Proceedings
A2 - Bubba, Tatiana A.
A2 - Gaburro, Romina
A2 - Gazzola, Silvia
A2 - Papafitsoros, Kostas
A2 - Pereyra, Marcelo
A2 - Schönlieb, Carola-Bibiane
PB - Springer
CY - Cham, Switzerland
T2 - 10th International Conference on Scale Space and Variational Methods in Computer Vision, SSVM 2025
Y2 - 18 May 2025 through 22 May 2025
ER -
