Groups
Category
Equivariance means that applying a transformation before a function is the same as applying a corresponding transformation after the function.
Data augmentation expands the training distribution by applying label-preserving transformations to inputs, which lowers overfitting and improves generalization.