Open Set Domain Adaptation Using Optimal Transport

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