Global Guarantees for Blind Demodulation with Generative Priors

Neural Information Processing Systems 

We study a deep learning inspired formulation for the blind demodulation problem, which is the task of recovering two unknown vectors from their entrywise multiplication.