Abstract
Abstract This paper explores the use of transfer learning in artificial neural networks modelling of Annular Aperture Arrays (AAAs) and Nanohole Arrays (NHAs). An AAA model trained on a larger dataset was used to enhance the performance of an NHA model through transfer learning. The NHA model with transfer learning achieved convergence speed an order of magnitude faster and reduced validation loss by 37% compared to the baseline NHA model. Comparative analysis shows that the optical responses are primarily influenced by the periodicity rather than other structural parameters. This study demonstrates the significant potential of transfer learning to improve accuracy and efficiency in modelling optical nanodevices, especially when data availability is limited.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call