Fig. 2: Proposed RPN for the design space of spectral response mapping. | Light: Science & Applications

Fig. 2: Proposed RPN for the design space of spectral response mapping.

From: Machine learning assisted plasmonic metascreen for enhanced broadband absorption in ultra-thin silicon films

Fig. 2

a Schematic of the RPN. b The decrease of training and validation loss during RPN training. c Statistical distribution of relative spectral error when applying the trained RPN on the test dataset, where the red dashed line shows the mean error. d Representative examples of predicted absorption spectra, where the solid blue and the dashed red curves indicate the target and RPN predicted results, respectively. The corresponding design parameters for (i–iii) are provided in the SI

Back to article page