A machine learning approach for designing surface plasmon resonance PCF-based sensors

This study explores the application of machine learning algorithms to optimize the geometry of the plasmonic layer in a surface plasmon resonance photonic crystal fiber sensor. By leveraging the simplicity of linear regression (LR) alongside the advanced predictive capabilities of the gradient boosted regression (GBR) algorithm, the proposed approach enables accurate prediction and optimization of the plasmonic layer’s configuration to achieve a desired spectral response. The integration of LR and GBR with computational simulations yielded impressive results, with an R² exceeding 0.97 across all analyzed variables. Moreover, the predictive accuracy demonstrated a remarkably low margin of error, ε < 10^-15. This combination of methods provides a robust and efficient pathway for optimizing sensor design, ensuring enhanced performance and reliability in practical applications.