Algorithms and Theory for Multiple-Source Adaptation
This work provides a full solution for multiple-source adaptation, which is a domain-specific problem in machine learning for handling distinct source domains, but it appears incremental as it builds on existing adaptation frameworks.
The paper tackles the multiple-source adaptation problem by developing new normalized solutions and algorithms with strong theoretical guarantees for cross-entropy and similar losses, and it reports that their algorithm outperforms competing approaches by producing a robust model that performs well on any target mixture distribution in experiments with real-world datasets.
This work includes a number of novel contributions for the multiple-source adaptation problem. We present new normalized solutions with strong theoretical guarantees for the cross-entropy loss and other similar losses. We also provide new guarantees that hold in the case where the conditional probabilities for the source domains are distinct. Moreover, we give new algorithms for determining the distribution-weighted combination solution for the cross-entropy loss and other losses. We report the results of a series of experiments with real-world datasets. We find that our algorithm outperforms competing approaches by producing a single robust model that performs well on any target mixture distribution. Altogether, our theory, algorithms, and empirical results provide a full solution for the multiple-source adaptation problem with very practical benefits.