Phase-limited quantization-aware training for diffractive deep neural networks
Yu Wang, Qi Sha, Qi Feng · Applied Optics · 2025
In recent years, all-optical diffractive deep neural networks (D 2 NNs) have demonstrated exceptional performances in many fields. Discretizing the grating height can reduce the complexity and enhance the network manufacturing efficiency. For this purpose, we propose a phase-limited quantization-aware training (PLQAT) method and construct an all-optical D 2 NN to discretize the network for the MNIST image classification task. Our results indicate that the PLQAT method improves the classification performance of the D 2 NN by 0.11–27.96% across different bit levels compared to the classical algorithm. Furthermore, we identified 3-bit quantization as the optimal choice, discretized the phase values in five layers to eight levels within [0,2 π ], and achieved a test accuracy of 96.22%. This method discretizes the height of gratings of D 2 NNs, effectively reducing the difficulty of grating etching while maintaining good network performance.