Adversarial Active Learning for Deep Networks: a Margin Based Approach | Read Paper on Bytez