Fig. 3: The pipeline of the Ψ-NN method.
From: Automatic network structure discovery of physics informed neural networks via knowledge distillation

Using Burgers equation as a sample. Three bold arrows represent the distillation-extraction-reconstruction process. a The whole field data was predicted by data-driven model. b The student network is trained with the structure generating method, using the teacher model’s output. c The structure extraction and reconstruction method is applied to find an optimal network structure for the Burgers equation modeling.