Phase Retrieval is a difficult ill posed inverse problem due to its highly non-convex nature. To overcome its difficulties Metzler et al (2018) proposed to use Deep learning based regularization.
The objective of this project is to understand and implement the proposed method and to evaluate its performance with respect to previously proposed techniques.
Metzler, C. A., Schniter, P., Veeraraghavan, A., & Baraniuk, R. G. (2018). prDeep: Robust Phase Retrieval with a Flexible Deep Network. [arXiv:1803.00212]