Open Set Domain Adaptation using Optimal Transport

Marwa Kechaou (1, 2, 3, 4) , Romain Hérault (1, 2, 3, 4) , Mokhtar Z. Alaya (1, 2, 3, 5) , Gilles Gasso (1, 2, 3, 4)

1 DocApp - LITIS - Equipe Apprentissage
2 LITIS - Laboratoire d'Informatique, de Traitement de l'Information et des Systèmes
3 NU - Normandie Université
4 INSA Rouen Normandie - Institut national des sciences appliquées Rouen Normandie
5 UNIROUEN - Université de Rouen Normandie

We present a 2-step optimal transport approach that performs a mapping from a source distribution to a target distribution. Here, the target has the particularity to present new classes not present in the source domain. The first step of the approach aims at rejecting the samples issued from these new classes using an optimal transport plan. The second step solves the target (class ratio) shift still as an optimal transport problem. We develop a dual approach to solve the optimization problem involved at each step and we prove that our results outperform recent state-of-the-art performances. We further apply the approach to the setting where the source and target distributions present both a label-shift and an increasing covariate (features) shift to show its robustness.

Domains: Statistics [stat]; Machine Learning [stat.ML]

Marwa Kechaou, Romain Hérault, Mokhtar Z. Alaya, Gilles Gasso. Open Set Domain Adaptation using Optimal Transport. European machine learning and data mining conference (ECML-PKDD), Sep 2020, Ghent, Belgium. ⟨hal-02959646⟩