Concise Explanations of Neural Networks using Adversarial Training