Self-adaptive weighting and sampling for physics-informed neural networks
This work addresses a domain-specific problem for researchers and practitioners using PINNs to solve PDEs, offering an incremental improvement by combining existing adaptive techniques.
The paper tackles the challenge of limited accuracy and efficiency in physics-informed neural networks (PINNs) for solving partial differential equations (PDEs) by introducing a hybrid adaptive sampling and weighting method, which consistently improves prediction accuracy and training efficiency compared to using either strategy alone.
Physics-informed deep learning has emerged as a promising framework for solving partial differential equations (PDEs). Nevertheless, training these models on complex problems remains challenging, often leading to limited accuracy and efficiency. In this work, we introduce a hybrid adaptive sampling and weighting method to enhance the performance of physics-informed neural networks (PINNs). The adaptive sampling component identifies training points in regions where the solution exhibits rapid variation, while the adaptive weighting component balances the convergence rate across training points. Numerical experiments show that applying only adaptive sampling or only adaptive weighting is insufficient to consistently achieve accurate predictions, particularly when training points are scarce. Since each method emphasizes different aspects of the solution, their effectiveness is problem dependent. By combining both strategies, the proposed framework consistently improves prediction accuracy and training efficiency, offering a more robust approach for solving PDEs with PINNs.