Class-Incremental Domain Adaptation
It addresses a practical challenge for machine learning systems needing to handle both domain shifts and new classes, though it appears incremental by combining existing paradigms.
The paper tackles the problem of adapting models to new target-domain classes under domain shift, proposing a method that outperforms existing domain adaptation and class-incremental approaches in this setting.
We introduce a practical Domain Adaptation (DA) paradigm called Class-Incremental Domain Adaptation (CIDA). Existing DA methods tackle domain-shift but are unsuitable for learning novel target-domain classes. Meanwhile, class-incremental (CI) methods enable learning of new classes in absence of source training data but fail under a domain-shift without labeled supervision. In this work, we effectively identify the limitations of these approaches in the CIDA paradigm. Motivated by theoretical and empirical observations, we propose an effective method, inspired by prototypical networks, that enables classification of target samples into both shared and novel (one-shot) target classes, even under a domain-shift. Our approach yields superior performance as compared to both DA and CI methods in the CIDA paradigm.