Stochastic parallel gradient descent optimization based on decoupling of the software and hardware

摘要

We classified the decoupled stochastic parallel gradient descent (SPGD)optimization model into two different types: software and hardware decouplingmethods. A kind of software decoupling method is then proposed and a kind ofhardware decoupling method is also proposed depending on the Shack-Hartmann(S-H) sensor. Using the normal sensor to accelerate the convergence ofalgorithm, the hardware decoupling method seems a capable realization ofdecoupled method. Based on the numerical simulation for correction of phasedistortion in atmospheric turbulence, our methods are analyzed and comparedwith basic SPGD model and also other decoupling models, on the aspects ofdifferent spatial resolutions, mismatched control channels and noise. Theresults show that the phase distortion can be compensated after tens iterationswith a strong capacity of noise tolerance in our model.