CrossTrainer: Practical Domain Adaptation with Loss Reweighting
This work addresses a practical bottleneck for users needing efficient domain adaptation with data from varying sources, though it appears incremental as it builds on existing loss reweighting techniques.
The paper tackles the problem of expensive hyperparameter tuning in domain adaptation by introducing CrossTrainer, a system that uses loss reweighting and optimizations to achieve high model accuracy across datasets while reducing training time compared to naive search.
Domain adaptation provides a powerful set of model training techniques given domain-specific training data and supplemental data with unknown relevance. The techniques are useful when users need to develop models with data from varying sources, of varying quality, or from different time ranges. We build CrossTrainer, a system for practical domain adaptation. CrossTrainer utilizes loss reweighting, which provides consistently high model accuracy across a variety of datasets in our empirical analysis. However, loss reweighting is sensitive to the choice of a weight hyperparameter that is expensive to tune. We develop optimizations leveraging unique properties of loss reweighting that allow CrossTrainer to output accurate models while improving training time compared to naive hyperparameter search.