Abstract
Abstract
In this paper, we consider the problem of phase retrieval under masked Fourier measurements. Specifically, we focus on a question raised by E Candès et al, namely, to reduce the sampling complexity of such problem in noiseless case. We provide a truncated Wirtinger flow (TWF) algorithm with resampling to solve the problem. This algorithm starts with an initial guess computed by some truncated spectral method, then proceeds by successively refining the estimation via an update rule that bears a strong resemblance to a gradient descent scheme. These careful selection rules provide better initial guess and descent direction. We prove that when the sampling complexity is O(n log n), we get some good initial solution. Furthermore, the sequence generated by the resampled TWF algorithm can linearly converge to the true signal
x
∈
C
n
with
ϵ
~
-accuracy up to a global phase, provided with the sampling complexity
O
(
log
(
1
ϵ
~
)
n
log
2
n
)
. To some degree, we improve the results proposed by E Candès et al, which demand O(n log4
n) samples.
Funder
National Natural Science Foundation of China
the fundamental research funds for the central universities
Subject
Applied Mathematics,Computer Science Applications,Mathematical Physics,Signal Processing,Theoretical Computer Science
Reference39 articles.
1. A flexible convex relaxation for phase retrieval;Bahmani;Electron. J. Stat.,2017
2. Painless reconstruction from magnitudes of frame coefficients;Balan;J. Fourier Anal. Appl.,2009
3. Phase retrieval from power spectra of masked signals;Bandeira;Inf. Infer.,2014
4. Solving systems of phaseless equations via Riemannian optimization with optimal sampling complexity;Cai,2018
5. Phase retrieval via matrix completion;Candès;SIAM J. Imaging Sci.,2013
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献