基于机器学习赋能的太赫兹超表面设计与应用OA

In recent years,the integration of terahertz metasurfaces with machine learning technologies has injected new vitality into this field.Leveraging efficient iterative rates and diverse network architectures,machine learning has been widely employed on one hand as surrogate models for full-wave numerical simulations and intelligent design of terahertz metasurfaces;on the other hand,terahertz metasurfaces provide a computing platform that operates at the speed of light for optical computing.This paper focuses on the application of machine learning in terahertz metasurfaces.It first elaborates on the design frameworks and principles based on deep neural networks and generative adversarial networks,revealing multi-dimensional parameter optimization mechanisms.Subsequently,taking the all-optical diffractive deep neural network as an example,it introduces the physical mechanisms by which metasurfaces implement machine learning algorithms and their derived applications.Finally,it prospects the development trends and application prospects of the integration between machine learning and terahertz metasurfaces.