Fig. 6: Holographic implementations of neural networks based on gradient descent optimizers.
From: Non-convex optimization for inverse problem solving in computer-generated holography
a DeepCGH149. b Holo-encoder148. c Learned hardware-in-the-loop phase retrieval151. d Neural holography with camera-in-the-loop training133. e Neural 3D holography for AR/VR display152. f Time-multiplexed neural holography153. References148,149 are reprinted with permission from © Optical Society of America. References133,151,152 are reprinted with permission from © ACM. Reference153 is reprinted with permission from the authors
