Reconfigurable design of a thermo-optically addressed liquid-crystal phase modulator by a neural network.

Abstract

We present a machine learning approach to program the light phase modulation function of an innovative thermo-optically addressed, liquid-crystal based, spatial light modulator (TOA-SLM). The designed neural network is trained with a little amount of experimental data and is enabled to efficiently generate prescribed low-order spatial phase distortions. These results demonstrate the potential of neural network-driven TOA-SLM technology for ultrabroadband and large aperture phase modulation, from adaptive optics to ultrafast pulse shaping.