Global convergence of gradient descent for phase retrieval

We propose a tensor-based criterion for benign landscape in phase retrieval and establish boundedness of gradient trajectories. This implies that gradient descent will converge to a global minimum for almost every initial point.

math.OC cs.LG Physics Computer Science Mathematics

Paper

Similar papers

Gradient Descent with Random Initialization: Fast Global Convergence for Nonconvex Phase RetrievalYuxin Chen, Yuejie Chi, Jianqing Fan et al.2018 · Mathematical programming · 262 citations In Library

On Gradient Descent Algorithm for Generalized Phase Retrieval ProblemJi Li, Tie Zhou2016 · arXiv: Optimization and Control · 2 citations In Library

Stochasticity helps to navigate rough landscapes: comparing gradient-descent-based algorithms in the phase retrieval problemFrancesca Mignacco, Pierfrancesco Urbani, Lenka Zdeborov'a2021 · Machine Learning: Science and Technology · 43 citations In Library

Convolutional Phase Retrieval via Gradient DescentQing Qu, Yuqian Zhang, Yonina C. Eldar et al.2017 · IEEE Transactions on Information Theory · 33 citations In Library

On the conditional equivalence of phase retrieval algorithmsJ. Schroeder, Andreas Döpp2026 · Machine Learning: Science and Technology · 0 citations In Library

Robust Gradient Descent for Phase RetrievalAlex Buna, Patrick Rebeschini2024 · International Conference on Artificial Intelligence and Statistics · 2 citations In Library

Online Stochastic Gradient Descent with Arbitrary Initialization Solves Non-smooth, Non-convex Phase RetrievalYan Shuo Tan, R. Vershynin2019 · Journal of machine learning research · 46 citations In Library

Phase Retrieval Meets Statistical Learning Theory: A Flexible Convex RelaxationS. Bahmani, J. Romberg2016 · International Conference on Artificial Intelligence and Statistics · 129 citations In Library

Alternating Phase Projected Gradient Descent with Generative Priors for Solving Compressive Phase RetrievalRakib Hyder, Viraj Shah, C. Hegde et al.2019 · IEEE International Conference on Acoustics, Speech, and Signal Processing · 49 citations In Library

Over-Parameterized Network Solves Phase Retrieval EffectivelyJi Li, Chao Wang2022 · IEEE International Conference on Acoustics, Speech, and Signal Processing · 0 citations In Library

Source of record

arXiv

44B

Welcome.

This panel sticks with you. Pick anything below and it opens right beside it, so you can dig through 60,000-plus records without ever losing your spot here.

Travel 44B

LibraryPapers, policy, standards, statute LabsEvery organization building AI ModelsIntelligence, price, and speed BenchmarksThe evaluation catalog IncidentsAI failures and harms SearchOne field across everything DashboardThe state of AI accountability in NY 44B RegistryThe Art. 44-B compliance portal