In this paper, we will solve the generalized phase retrieval (PR) problem over a network, where each agent only has a subset of the measurements. The problem is formulated as minimizing the squared loss between the measurements and linear sensing intensity. To solve the problem in a distributed setting, an algorithm named distributed Wirtinger flow (DWF) is proposed. Theoretical analyses show that the proposed DWF algorithm converges to the (approximate) KKT points of the original problem globally in a sublinear rate. The performance of the DWF algorithm is numerically compared with the state-of-the-art method. Simulation results show that DWF is able to recover a high-quality solution for the original PR problem.
|Original language||English (US)|
|Title of host publication||Conference Record of the 52nd Asilomar Conference on Signals, Systems and Computers, ACSSC 2018|
|Editors||Michael B. Matthews|
|Publisher||IEEE Computer Society|
|Number of pages||5|
|State||Published - Feb 19 2019|
|Event||52nd Asilomar Conference on Signals, Systems and Computers, ACSSC 2018 - Pacific Grove, United States|
Duration: Oct 28 2018 → Oct 31 2018
|Name||Conference Record - Asilomar Conference on Signals, Systems and Computers|
|Conference||52nd Asilomar Conference on Signals, Systems and Computers, ACSSC 2018|
|Period||10/28/18 → 10/31/18|
Bibliographical noteFunding Information:
This work was supported in part by the University of Minnesota and the Hong Kong University of Science and Technology via bilateral collaboration funds, and in part by the Sponsorship Scheme for Targeted Strategic Partnership FP602 of the Hong Kong University of Science and Technology. The work of Z. Zhao was supported by the Hong Kong PhD Fellowship Scheme.
© 2018 IEEE.
- Quadratic systems
- decentralized optimization
- distributed learning
- nonconvex optimization
- statistical learning over networks