AdversarialAuto

Data augmentation has emerged asaneffectivedata pre-processing ordata transformation step to mitigateoverfitting [31],toencourage local smoothness [57],andtoimprovegeneralization [6]in machine learning pipelines such as deep neural networks.