报告题目：Phaseless sampling and reconstruction

报 告 人：成诚（杜克大学 & SAMSI 博士后）

报告时间：2018年6月14日 10:00-11:00

报告地点：数统院307学术报告厅

报告摘要：

In this talk, we consider whether it is possible to determine the signal $f$ in a shift-invariant space, up to a sign, from its magnitude measurements either on the Euclidean space or on a discrete set. This is closely related to the phase retrieval problem of real-valued signals in an infinite-dimensional space. For the signal living in a shift-invariant space, we introduce an undirected graph associated with it and use connectivity of the graph to characterize its phase retrievability. We also propose a stable algorithm to reconstruct the signal from its noisy phaseless samples taken on a discrete set. The proposed algorithm provides a suboptimal approximation to the original signal in magnitude measurements.