Towards Accurate Model Selection in Deep Unsupervised Domain Adaptation

2026-06-03

AI Summary

This work addresses the lack of reliable, label-free model selection methods in deep unsupervised domain adaptation, which hinders fair algorithmic comparison. To this end, the paper proposes Deep Embedding Validation (DEV), the first approach capable of providing an unbiased and low-variance estimate of target risk without access to target-domain labels. DEV integrates adapted feature representations into the validation process and leverages control variates to effectively reduce estimation variance. Theoretical analysis and extensive experiments demonstrate that DEV significantly outperforms existing model selection strategies in both effectiveness and stability, offering a principled solution for hyperparameter tuning and model evaluation in unsupervised domain adaptation settings.

Abstract

Deep unsupervised domain adaptation (Deep UDA) methods successfully leverage rich labeled data in a source domain to boost the performance on related but unlabeled data in a target domain. However, algorithm comparison is cumbersome in Deep UDA due to the absence of accurate and standardized model selection method, posing an obstacle to further advances in the field. Existing model selection methods for Deep UDA are either highly biased, restricted, unstable, or even controversial (requiring labeled target data). To this end, we propose \textit{Deep Embedded Validation} (\textbf{DEV}), which embeds adapted feature representation into the validation procedure to obtain unbiased estimation of the target risk with bounded variance. The variance is further reduced by the technique of control variate. The efficacy of the method has been justified both theoretically and empirically.

Problem

• Deep Unsupervised Domain Adaptation
• Model Selection
• Target Risk Estimation
• Domain Adaptation
• Unbiased Validation

Innovation

• Deep Embedded Validation
• Unsupervised Domain Adaptation
• Model Selection
• Target Risk Estimation
• Control Variate