Domain Adaptation without Model Transferring
This addresses privacy risks in data-critical scenarios by enabling domain adaptation without model transfer, though it appears incremental as it builds on existing UDA methods.
The paper tackles the problem of domain adaptation without transferring source models to protect data privacy, proposing a method that refines information from the source model and uses distributionally adversarial training, achieving feasibility demonstrated on benchmarks like Digit-Five and Office-31.
In recent years, researchers have been paying increasing attention to the threats brought by deep learning models to data security and privacy, especially in the field of domain adaptation. Existing unsupervised domain adaptation (UDA) methods can achieve promising performance without transferring data from source domain to target domain. However, UDA with representation alignment or self-supervised pseudo-labeling relies on the transferred source models. In many data-critical scenarios, methods based on model transferring may suffer from membership inference attacks and expose private data. In this paper, we aim to overcome a challenging new setting where the source models cannot be transferred to the target domain. We propose Domain Adaptation without Source Model, which refines information from source model. In order to gain more informative results, we further propose Distributionally Adversarial Training (DAT) to align the distribution of source data with that of target data. Experimental results on benchmarks of Digit-Five, Office-Caltech, Office-31, Office-Home, and DomainNet demonstrate the feasibility of our method without model transferring.