Fetching the paper…

Regularising for invariance to data augmentation improves supervised learning · Around