Causally Inspired Regularization Enables Domain General Representations
This work addresses domain generalization for machine learning practitioners, offering a novel causal approach to reduce spurious correlations, though it builds incrementally on existing causal graph literature.
The paper tackles the problem of identifying domain-general feature representations in predictive settings by categorizing causal graphs into two groups and proposing a regularization framework for cases where standard empirical risk minimization fails. The method outperforms state-of-the-art approaches in average and worst-domain transfer accuracy on synthetic and real-world data.
Given a causal graph representing the data-generating process shared across different domains/distributions, enforcing sufficient graph-implied conditional independencies can identify domain-general (non-spurious) feature representations. For the standard input-output predictive setting, we categorize the set of graphs considered in the literature into two distinct groups: (i) those in which the empirical risk minimizer across training domains gives domain-general representations and (ii) those where it does not. For the latter case (ii), we propose a novel framework with regularizations, which we demonstrate are sufficient for identifying domain-general feature representations without a priori knowledge (or proxies) of the spurious features. Empirically, our proposed method is effective for both (semi) synthetic and real-world data, outperforming other state-of-the-art methods in average and worst-domain transfer accuracy.