Journal Home Online First Current Issue Archive For Authors Journal Information 中文版

Frontiers of Information Technology & Electronic Engineering >> 2022, Volume 23, Issue 12 doi: 10.1631/FITEE.2200284

Dual collaboration for decentralized multi-source domain adaptation

Affiliation(s): College of Intelligence and Computing, Tianjin University, Tianjin 300350, China; Tianjin Key Lab of Machine Learning, Tianjin University, Tianjin 300350, China; less

Received: 2022-06-29 Accepted: 2022-12-14 Available online: 2022-12-14

Next Previous

Abstract

The goal of decentralized is to conduct unsupervised in a scenario. The challenge of is that the source domains and target domain lack cross-domain collaboration during training. On the unlabeled target domain, the target model needs to transfer supervision knowledge with the collaboration of source models, while the domain gap will lead to limited adaptation performance from source models. On the labeled source domain, the source model tends to overfit its domain data in the scenario, which leads to the problem. For these challenges, we propose dual collaboration for decentralized by training and aggregating the local source models and local target model in collaboration with each other. On the target domain, we train the local target model by distilling supervision knowledge and fully using the unlabeled target domain data to alleviate the problem with the collaboration of local source models. On the source domain, we regularize the local source models in collaboration with the local target model to overcome the problem. This forms a dual collaboration between the decentralized source domains and target domain, which improves the domain adaptation performance under the scenario. Extensive experiments indicate that our method outperforms the state-of-the-art methods by a large margin on standard datasets.

Related Research