Entropic Projection Alignment: Estimating, Explaining, and Improving Model Performance Under Distribution Shift
This work addresses the critical problem of maintaining and improving model performance under distribution shift for machine learning practitioners, offering a unified framework for estimation, explanation, and adaptation.
This paper introduces Entropic Projection Alignment (EPA), a method to estimate, explain, and improve model performance under distribution shift. EPA aligns source and target distributions by matching selected moments while minimizing KL divergence, leading to a closed-form solution for importance weights. Experiments show EPA consistently outperforms state-of-the-art baselines with improved computational efficiency.
We propose a unified framework for addressing three key challenges of distribution shift: (1) estimating a model's performance on an unlabeled target domain, (2) explaining the shift by identifying the features responsible, and (3) improving the target domain performance. Our method, Entropic Projection Alignment (EPA), aligns the source distribution to the target by matching carefully selected moments while simultaneously minimising the KL divergence from the source. This formulation yields a unique closed-form solution for importance weights, achieving robustness through implicit variance control. Drawing on domain adaptation theory, we establish that moment matching is sufficient for reliable estimation and adaptation, avoiding the need for full density ratio recovery. Extensive experiments, together with strong theoretical guarantees, demonstrate that EPA consistently outperforms state-of-the-art baselines while offering substantial computational efficiency.