Physic Aware

"Physics-aware" methods integrate physical principles into machine learning models to improve accuracy, efficiency, and generalizability in scientific and engineering applications. Current research focuses on incorporating physics knowledge into various neural network architectures, including transformers and physics-informed neural networks (PINNs), often using techniques like incorporating physical constraints into loss functions or pretraining models on physics-based datasets. This approach enhances model performance, particularly in scenarios with limited data or complex dynamics, leading to improved predictions in diverse fields such as robotics, fluid dynamics, and materials science.