Quantifying the Preferential Direction of the Model Gradient in Adversarial Training With Projected Gradient Descent

Re-assign community