Adversarial Unlearning: Reducing Confidence Along Adversarial Directions

Computer Science

• Adversarial Machine Learning 100%
• Data Augmentation 33%
• Lines of Code 33%
• maximum-likelihood 33%
• Regularization 33%
• Supervised Learning 33%
• Training Data 100%

Keyphrases

• Label Smoothing 66%
• Low Data Regimes 33%
• Mixup Training 33%
• Spurious Features 33%