When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations

Chen, Xiangning, Hsieh, Cho-Jui, Gong, Boqing · arXiv · 2021