Physic Based Feature

Physics-based features are increasingly integrated into machine learning models to improve accuracy, generalizability, and interpretability across diverse scientific and engineering domains. Current research focuses on incorporating these features into various architectures, including deep neural networks and graph convolutional networks, often employing techniques like physics-guided feature extraction and domain adaptation to enhance model performance. This approach addresses limitations of purely data-driven models, particularly in scenarios with limited data or significant domain shifts, leading to more robust and reliable predictions in applications ranging from structural engineering and power systems to tokamak disruption prediction and gaze estimation. The resulting models offer improved performance and provide valuable insights into the underlying physical processes.