CODA: A Correlation-Oriented Disentanglement and Augmentation Modeling Scheme for Better Resisting Subpopulation Shifts

Neural Information Processing Systems 

Data-driven models learned often struggle to generalize due to widespread subpopulation shifts, especially the presence of both spurious correlations and group imbalance (SC-GI).

Similar Docs  Excel Report  more

TitleSimilaritySource
None found