Fig. 4: Basic principle of using Physics-Informed Neural Networks for metasurface design. | Light: Science & Applications

Fig. 4: Basic principle of using Physics-Informed Neural Networks for metasurface design.

From: Recent advances in metasurface design and quantum optics applications with machine learning, physics-informed neural networks, and topology optimization methods

Fig. 4

Detailed steps are demonstrated in the flowchart, including define the problem, create the training data, define physics principle and loss function, define the PINN architecture, validate the network, train backpropagation networks, input desired EM response, iterate the design process, and obtain desired metasurface designs

Back to article page