27
12

Exact Minimax Estimation for Phase Synchronization

Abstract

We study the phase synchronization problem with measurements Y=zzH+σWCn×nY=z^*z^{*H}+\sigma W\in\mathbb{C}^{n\times n}, where zz^* is an nn-dimensional complex unit-modulus vector and WW is a complex-valued Gaussian random matrix. It is assumed that each entry YjkY_{jk} is observed with probability pp. We prove that the minimax lower bound of estimating zz^* under the squared 2\ell_2 loss is (1o(1))σ22p(1-o(1))\frac{\sigma^2}{2p}. We also show that both generalized power method and maximum likelihood estimator achieve the error bound (1+o(1))σ22p(1+o(1))\frac{\sigma^2}{2p}. Thus, σ22p\frac{\sigma^2}{2p} is the exact asymptotic minimax error of the problem. Our upper bound analysis involves a precise characterization of the statistical property of the power iteration. The lower bound is derived through an application of van Trees' inequality.

View on arXiv
Comments on this paper