Using WGAN for Improving Imbalanced Classification Performance.
Snehal Bhatia, Rozenn Dahyot · Maynooth University ePrints and eTheses Archive (Maynooth University) · 2019
This paper investigates data synthesis with a Generative Adversarial Network (GAN) for augmenting the amount of data used for training classifiers (in supervised learning) to compensate for class imbalance (when the classes are not represented equally by the same number of training samples). Our data synthesis approach with GAN is compared with data augmentation in the context of image classification. Our experimental results show encouraging results in comparison to standard data augmentation schemes based on image transforms.