Explainable AI in computational optics: interpreting material property relationships

Explainable AI plays a crucial role in material science, particularly for interpreting material-property relationships. Coupled with existing optical theory codes, explainable AI elucidates the underlying physics of material behavior in optical materials. An analysis framework integrates various explain-ability techniques to interpret the interdependence of multiple optical and material properties. The approach is applied to glasses, polymers, and multicomponent composites. Fine-grained materials classification improves the accuracy of optical-property prediction. In glasses, the fundamental composition–property relationship can be elucidated. The segregation of elemental properties accounts for the variation in refractive-index behavior. Formulas with guiding physical insights are automatically discovered. Interpreted AI enables data-driven material design for mould-production-quality control. The proposed method offers an interpretability-driven development paradigm, representing the next step toward explainable, trustworthy material design, which bridges prior physics rules and experimental observations in computational optics.

Перевод пока недоступен