CFSM: A Novel Causal Feature Selection Module for Two-Dimensional Out-of-Distribution Generalization.

In real-world scenarios, training and test data are often collected in diverse settings, leading to domain shifts arising from evolving environments and selection bias. While causality-inspired methods have shown promising results in tackling the out…

Preview only. Read the full abstract at the source