arXiv · 1805.12018
Generalizing to Unseen Domains via Adversarial Data Augmentation
Abstract
We are concerned with learning models that generalize well to different \emph{unseen} domains. We consider a worst-case formulation over data distributions that are near the source domain in the feature space. Only using training data from a single source distribution, we propose an iterative procedure that augments the dataset with examples from a fictitious target domain that is "hard" under the current model. We show that our iterative scheme is an adaptive data augmentation method where we append adversarial examples at each iteration. For softmax losses, we show that our method is a data-dependent regularization scheme that behaves differently from classical regularizers that regularize towards zero (e.g., ridge or lasso). On digit recognition and semantic segmentation tasks, our method learns models improve performance across a range of a priori unknown target domains.
Explore related subjects
Keep this discovery
Riccardo Volpi, Hongseok Namkoong, Ozan Sener, John Duchi, Vittorio Murino, Silvio Savarese. 2018-05-30. Generalizing to Unseen Domains via Adversarial Data Augmentation. https://arxiv.org/abs/1805.12018
Cite the original work for its findings. Save a collection to share your selection of sources.