Van der Pol-informed Neural Networks for Multi-step-ahead Forecasting of Extreme Climatic Events.
Advances in Neural Information Processing Systems (NeurIPS) - AI4Science., 2023
Physics-informed neural networks (PINNs) integrate domain knowledge with data-driven learning, offering a powerful framework for modeling complex dynamical systems. This work proposes VPINN, a Van der Pol-informed neural network designed to capture extreme nonlinear dynamics in climatic systems using physics-based loss functions. The approach enhances both data efficiency and physical consistency.
<img src="/images/VPINN_Model_image.jpg" alt=VPINN_Model_image" style="width: 90%; max-width: 800px;">
Architecture of the VPINN model.