Fig. 7: Phase-retrieval problems with Poisson or systematic errors.
From: Phase retrieval via gain-based photonic XY-Hamiltonian optimization
These problems were generated from random complex-valued samples with L = 5 phase filters and with Poisson noise or systematic errors added to observation vectors. For each problem, the gain-based solver evolved to t = 1000, and the RRR ran for 10,000 iterations. a Phase-retrieval error (RSE) produced by the RRR method (dashed line) and the gain-based system (solid line) as a function of Poisson noise in the measured amplitudes. b RSE produced by RRR and the gain-based methods as a function of systematic errors introduced into the measured amplitudes. In both panels, vertical error bars are the standard deviations in RSE across 20 different CDP phase-retrieval problems generated from random complex-valued sample vectors. These sample vectors had amplitudes uniformly randomly distributed in [0, 1) and phases uniformly randomly distributed in [0, 2π). See “Fig 7a” and “Fig. 7b” tables in supplementary data.
