Active Domain Adaptation
A machine learning paradigm combining active learning with domain adaptation. It permits querying a small number of target-domain sample labels to improve unsupervised domain adaptation performance while reducing annotation cost.
Timeline (last 90 days)
Active domain adaptation (ADA) allows querying the labels of a small number of target-domain samples, trading annotation cost for performance improvement.
Traditional active domain adaptation methods optimize for average accuracy, which may cause performance degradation on certain difficult classes and affect model robustness.
All Facts (2)
Active domain adaptation (ADA) allows querying the labels of a small number of target-domain samples, trading annotation cost for performance improvement.
50%UnverifiedTraditional active domain adaptation methods optimize for average accuracy, which may cause performance degradation on certain difficult classes and affect model robustness.
50%