Continual Domain Adversarial Adaptation via Double-Head Discriminators
This addresses a specific bottleneck in continual domain adaptation for machine learning practitioners, though it appears incremental.
The paper tackles the challenge of domain adversarial adaptation in continual learning where previous source domain data is inaccessible, by proposing a double-head discriminator algorithm that reduces empirical estimation error of H-divergence. Experiments show over 2% improvement on target domain adaptation tasks while mitigating source domain forgetting.
Domain adversarial adaptation in a continual setting poses a significant challenge due to the limitations on accessing previous source domain data. Despite extensive research in continual learning, the task of adversarial adaptation cannot be effectively accomplished using only a small number of stored source domain data, which is a standard setting in memory replay approaches. This limitation arises from the erroneous empirical estimation of $\gH$-divergence with few source domain samples. To tackle this problem, we propose a double-head discriminator algorithm, by introducing an addition source-only domain discriminator that are trained solely on source learning phase. We prove that with the introduction of a pre-trained source-only domain discriminator, the empirical estimation error of $\gH$-divergence related adversarial loss is reduced from the source domain side. Further experiments on existing domain adaptation benchmark show that our proposed algorithm achieves more than 2$\%$ improvement on all categories of target domain adaptation task while significantly mitigating the forgetting on source domain.