Аннотация
Artificial intelligence (AI) is used for the photonic inverse design of multilayered cylindrical nanorods that exhibit maximal polarization selectivity in the way they scatter the visible light. The optimal setups are obtained from a neural network whose training dataset contains only designs with significantly lower performance. The reported particles can be incorporated into on-chip systems, serving a wide spectrum of photonic applications from ultra-efficient sensing and imaging to secure optical communications and sharp filtering.