Passive Symmetry

Passive symmetry in machine learning focuses on exploiting inherent data invariances to improve model efficiency, robustness, and generalization. Current research emphasizes developing neural network architectures, such as equivariant graph neural networks and symmetry-enforcing neural networks, that explicitly incorporate these symmetries, often through tensor-based operations or ensemble methods. This approach is proving valuable across diverse applications, including material science (constitutive modeling), path planning, and even large language model evaluation, by enhancing data efficiency and improving performance, particularly in scenarios with limited data or complex symmetries.