Probability-Density-Based Deep Learning Paradigm for the Fuzzy Design of Functional Metastructures
Architecture of the proposed neural network. The input of the PDN is the targeted transmission spectrum, while the output is a mixture Gaussian that provides a probabilistic sampling to derive a plausible metastructure with similar transmission. Here, the mixture Gaussian is linearly superposed by individual Gaussians in the output PDN layer that are characterized by the parameters of mixing coefficient , mean , and deviation . Three output Gaussians are plotted as an example. The local maximum in the mixture Gaussian is mapping to an inversely designed structures with transmission spectra mostly close to the target one.
Please note that the Table of Contents page is currently not displaying content. We are working to fix this issue as soon as possible and apologize for the inconvenience.