2023-06-252023-06-25https://tore.tuhh.de/handle/11420/15782Neural networks are applied to investigate structure-property relations of structured optical materials. Such properties as effective medium parameters, photonic band diagrams and band gaps are considered. Methods are developed to address the direct and inverse problems, which are prediction of properties from structure and proposal of structure for required properties, correspondingly. Also, an optimization problem is solve, where topology is changed to maximize required properties.