Alternating Phase Projected Gradient Descent with Generative Priors for Solving Compressive Phase Retrieval

  • 2019-03-07 03:03:14
  • Rakib Hyder, Viraj Shah, Chinmay Hegde, M. Salman Asif
  • 15


The classical problem of phase retrieval arises in various signal acquisitionsystems. Due to the ill-posed nature of the problem, the solution requiresassumptions on the structure of the signal. In the last several years, sparsityand support-based priors have been leveraged successfully to solve thisproblem. In this work, we propose replacing the sparsity/support priors withgenerative priors and propose two algorithms to solve the phase retrievalproblem. Our proposed algorithms combine the ideas from AltMin approach fornon-convex sparse phase retrieval and projected gradient descent approach forsolving linear inverse problems using generative priors. We empirically showthat the performance of our method with projected gradient descent is superiorto the existing approach for solving phase retrieval under generative priors.We support our method with an analysis of sample complexity with Gaussianmeasurements.


Quick Read (beta)

loading the full paper ...