AI-driven approaches in electromagnetic metamaterials design and application: a review

At present, the increasingly sophisticated functional demands for electromagnetic metamaterials pose significant challenges to conventional design approaches reliant on numerical simulation and empirical knowledge, particularly in terms of computational efficiency and design capability. The integration of artificial intelligence has revitalized metamaterials research, enabling rapid forward performance prediction from structural parameters, inverse design from target properties to structural configurations, as well as adaptive functional enhancement and physical mechanism interpretation. This review examines two dominant pathways in AI-assisted electromagnetic metamaterial design and application: forward prediction coupled with inverse design, and adaptive control integrated with physical insight. We summarize recent advances in deep learning and generative models that enhance design efficiency, generalization capability, and interpretability of AI-driven metamaterial models. Furthermore, we explore emerging trends and persistent challenges in developing intelligent, self-adaptive, and multiphysics-coupled metamaterial systems powered by artificial intelligence.