Equivariance and Augmentation for Bayesian Neural Networks
Researchers from the University of Cambridge and the University of Oxford have derived conditions under which data augmentation yields exact equivariance in Bayesian neural networks trained with varia…