Margin-aware Adversarial Domain Adaptation with Optimal Transport

Sofien Dhouib (1) , Ievgen Redko (2) , Carole Lartizien (1)

1 Images et Modèles
2 LHC - Laboratoire Hubert Curien

In this paper, we propose a new theoretical analysis of unsupervised domain adaptation (DA) that relates notions of large margin separation, ad-versarial learning and optimal transport. This analysis generalizes previous work on the subject by providing a bound on the target margin violation rate, thus reflecting a better control of the quality of separation between classes in the target domain than bounding the misclassification rate. The bound also highlights the benefit of a large margin separation on the source domain for adaptation and introduces an optimal transport (OT) based distance between domains that has the virtue of being task-dependent, contrary to other approaches. From the obtained theoretical results, we derive a novel algorithmic solution for domain adaptation that introduces a novel shallow OT-based adversarial approach and outperforms other OT-based DA baselines on several simulated and real-world classification tasks.

Domaines: Probabilités [math.PR]; Intelligence artificielle [cs.AI]; Statistiques [math.ST]

Sofien Dhouib, Ievgen Redko, Carole Lartizien. Margin-aware Adversarial Domain Adaptation with Optimal Transport. Thirty-seventh International Conference on Machine Learning, Jul 2020, Vienne, Austria. ⟨hal-02900715⟩