Collaborative Semantic Aggregation and Calibration for Separated Domain Generalization

Domain generalization (DG) aims to learn from multiple known source domains a model that can generalize well to unknown target domains. The existing DG methods usually rely on shared multi-source data fusion for generalizable model training. However, tremendous data is distributed across lots of places nowadays that can not be shared due to privacy policies, especially in some crucial areas like finance and medical care. A dilemma is thus raised between real-world data privacy protection and simultaneous multi-source semantic learning with the shared data. In this paper, we investigate a separated domain generalization task with separated source datasets that can only be used locally, which is vital for real-world privacy protection. We propose a novel solution called Collaborative Semantic Aggregation and Calibration (CSAC) to enable this challenging task. To fully absorb multi-source semantic information while avoiding unsafe data fusion, we first conduct data-free semantic aggregation by fusing the models trained on the separated domains layer-by-layer. To address semantic dislocation caused by domain shift, we further design cross-layer semantic calibration with an attention mechanism to align each semantic level and enhance domain invariance. We unify multi-source semantic learning and alignment in a collaborative way by repeating the semantic aggregation and calibration alternately, keeping each dataset localized, and privacy is thus carefully protected. Extensive experiments show the significant performance of our method in addressing this challenging task, which is even comparable to the previous DG methods with shared data.