Skip to content
DraconiticResearch
Search Sources API
181,686 studies31 sourcesUpdated 50 min ago
PreprintarXiv licencearXiv

Deeply unrolled alternating projections for phase retrieval

  • Jiayi Yao
  • Vasileios Charisopoulos
  • Rebecca Willett
Open at sourcePDFDOICite

Abstract

Deep unrolling methods for inverse problems augment traditional reconstruction algorithms with learned priors to improve their performance. These methods have been developed under the implicit assumption that the geometry of the measurement operator does not change between training and deployment; when this is no longer the case, the quality of obtained reconstructions can degrade significantly. In this paper, we argue that this is not an inherent limitation of the deep unrolling paradigm and can be mitigated by a careful choice of the underlying iterative algorithm. We propose a deep unrolling framework for the phase retrieval problem that augments the alternating projections method with a learned denoising step. Our method obtains high-quality reconstructions without the need for retraining in settings where conventional model-based alternatives struggle, such as when the dynamic range of the measurement operator changes significantly at test time.

Related

  1. Acceleration of the Method of Alternating Projections on subspaces by GMRES
    Luis Manuel Hernández-Ramos, Maryam Calles-Cabriles, Gorka Emaldi-Porco
    arXivPreprintarXiv licenceFull textNumerical analysis & scientific computingarXiv
  2. Volume Sampling and Spectral Equalization in Randomized Alternating Projections
    Alireza Entezari, Arunava Banerjee, Leila Kalantari
    arXivPreprintCC BY-NC-ND 4.0Full textNumerical analysis & scientific computingarXiv
  3. A new convergence proof for an iterative scheme for elliptic Cauchy problems based on alternating orthogonal projections
    Abhijit Guchhait, Adrian Muntean, Martin Lind
    arXivPreprintCC BY 4.0Full textAnalysis & PDEsNumerical analysis & scientific computingarXiv
  4. Polynomial Acceleration of Alternating Projections: Chebyshev Acceleration, Convergence Rates, and Retractions
    Dengyu Zheng, Shixiang Chen
    arXivPreprintarXiv licenceFull textNumerical analysis & scientific computingOptimisation & controlarXiv
  5. Adaptive Bregman Alternating Projections for Feasible Gromov-Wasserstein Learning
    Aoran Zhang, César A. Uribe
    arXivPreprintCC BY 4.0Full textOptimisation & controlarXiv

Cite

BibTeXRIS
BibTeX
@misc{yao2026deeply,
  title = {{Deeply unrolled alternating projections for phase retrieval}},
  author = {Jiayi Yao and Vasileios Charisopoulos and Rebecca Willett},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.17316},
  eprint = {2610.17316},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.17316},
}
RIS
TY  - UNPB
TI  - Deeply unrolled alternating projections for phase retrieval
AU  - Jiayi Yao
AU  - Vasileios Charisopoulos
AU  - Rebecca Willett
PY  - 2026
DA  - 2026/10/12
PB  - arXiv
DO  - 10.48550/arxiv.2610.17316
UR  - https://arxiv.org/abs/2610.17316
AB  - Deep unrolling methods for inverse problems augment traditional reconstruction algorithms with learned priors to improve their performance. These methods have been developed under the implicit assumption that the geometry of the measurement operator does not change between training and deployment; when this is no longer the case, the quality of obtained reconstructions can degrade significantly. In this paper, we argue that this is not an inherent limitation of the deep unrolling paradigm and can be mitigated by a careful choice of the underlying iterative algorithm. We propose a deep unrolling framework for the phase retrieval problem that augments the alternating projections method with a learned denoising step. Our method obtains high-quality reconstructions without the need for retraining in settings where conventional model-based alternatives struggle, such as when the dynamic range of the measurement operator changes significantly at test time.
ER  -