Improving Out-of-Distribution Generalization by Adversarial Training with Structured Priors

Open in new window