Meta-Optics-Based Diffractive Learning Machine for Angle-Multiplexed Multi-Task Processing

This paper presents an experimental demonstration of a large-scale metasurface-based diffractive learning machine for angle-multiplexed multi-task processing. The system utilizes a metasurface-based reflection cavity to construct a five-layer diffractive neural network. Diffractive neural networks offer advantages of low latency, low energy consumption, and massive parallelism in optical computing applications.