Physics-informed neural network (PINN) is a data-driven solver for differential equations, providing a unified framework to address both forward and inverse problems. However, the complexity of the objective function often leads to training failures, especially when solving high-frequency and multi-scale problems. The authors propose using transfer learning to enhance the robustness and convergence of training PINN, starting from low-frequency problems and gradually approaching high-frequency problems. This method requires fewer data points and less training time and can effectively train PINN to approximate solutions from low-frequency problems to high-frequency problems without increasing network parameters.
Publication date: 8 Jan 2024
Project Page: https://arxiv.org/abs/2401.02810
Paper: https://arxiv.org/pdf/2401.02810
