Active learning using Generative Adversarial Networks for improving generalization and avoiding distractor points

Heechul Lim, Kang-Wook Chon, Min-Soo Kim

Abstract: Highlights•Reasonable active learning criteria and an effective GAN-based training strategy.•ALGAD both improves generalization and avoids distractor points.•ALGAD performs well with any number of distractor points.•Relationship between avoiding distractor points and improving generalization.•ALGAD performs better across various datasets than existing active learning methods.