Fig. 4
From: Intelligent designs in nanophotonics: from optimization towards inverse creation
A neural network (NN) for meta-atoms. (a) A simple structure of NN with an input of X, an output of Y and two hidden layers of X(l−1) and X(l), respectively. (b) A magnified view of one artificial neuron layer to perform nonlinear transformation from layer X(l−1) to layer X(l). (c) The set-up and training process of NN-based model for meta-atoms. The procedures are composed of four labeled steps, starting from raw data generation, model training and validation, model test, to new applications