Stein Discrepancy for Unsupervised Domain Adaptation

TL;DR

This paper introduces a novel unsupervised domain adaptation method using Stein discrepancy, which is more effective with scarce target data and supports flexible modeling, outperforming prior methods on benchmarks.

Contribution

The paper proposes a Stein discrepancy-based UDA framework with kernelized and adversarial forms, suitable for low-data target regimes, and provides theoretical guarantees.

Findings

Outperforms prior UDA methods with limited target data

Supports flexible modeling of target distribution (Gaussian, GMM, VAE)

Provides theoretical bounds on target error and convergence rate

Abstract

Unsupervised domain adaptation (UDA) aims to improve model performance on an unlabeled target domain using a related, labeled source domain. A common approach aligns source and target feature distributions by minimizing a distance between them, often using symmetric measures such as maximum mean discrepancy (MMD). However, these methods struggle when target data is scarce. We propose a novel UDA framework that leverages Stein discrepancy, an asymmetric measure that depends on the target distribution only through its score function, making it particularly suitable for low-data target regimes. Our proposed method has kernelized and adversarial forms and supports flexible modeling of the target distribution via Gaussian, GMM, or VAE models. We derive a generalization bound on the target error and a convergence rate for the empirical Stein discrepancy in the two-sample setting. Empirically, our method consistently outperforms prior UDA approaches under limited target data across multiple benchmarks.