Post-Training Augmentation Invariance

Keenan Eikenberry, Lizuo Liu, Yoonsang Lee

This work develops a framework for post-training augmentation invariance, in which our goal is to add invariance properties to a pretrained network without altering its behavior on…