Evidential Graph Contrastive Alignment for Source-Free Blending-Target Domain Adaptation
This addresses a realistic domain adaptation scenario for machine learning applications, but it is incremental as it builds on existing DA methods.
The paper tackles the problem of Source-Free Blending-Target Domain Adaptation, where source data is unavailable and multiple target domains are mixed without labels, by proposing Evidential Contrastive Alignment to decouple domains and handle noisy pseudo-labels, achieving considerable gains and comparable results to methods with domain labels or source data.
In this paper, we firstly tackle a more realistic Domain Adaptation (DA) setting: Source-Free Blending-Target Domain Adaptation (SF-BTDA), where we can not access to source domain data while facing mixed multiple target domains without any domain labels in prior. Compared to existing DA scenarios, SF-BTDA generally faces the co-existence of different label shifts in different targets, along with noisy target pseudo labels generated from the source model. In this paper, we propose a new method called Evidential Contrastive Alignment (ECA) to decouple the blending target domain and alleviate the effect from noisy target pseudo labels. First, to improve the quality of pseudo target labels, we propose a calibrated evidential learning module to iteratively improve both the accuracy and certainty of the resulting model and adaptively generate high-quality pseudo target labels. Second, we design a graph contrastive learning with the domain distance matrix and confidence-uncertainty criterion, to minimize the distribution gap of samples of a same class in the blended target domains, which alleviates the co-existence of different label shifts in blended targets. We conduct a new benchmark based on three standard DA datasets and ECA outperforms other methods with considerable gains and achieves comparable results compared with those that have domain labels or source data in prior.