adversarial training

A technique in machine learning where a model is trained using both normal and adversarially perturbed inputs. This aims to improve the model's robustness against adversarial attacks and enhances its generalization to unseen data.

14 papers