Multi-source domain adaptation with graph embedding and adaptive label prediction

作者：

Highlights：

• Moment matching and graph embedding are leveraged in a unified framework for subspace learning.

• Clustering and structural risk minimization (SRM) are jointly optimized to update soft labels for the target domain.

• Extensive experiments on both multi-source and single-source domain adaptation show the superiority of our method compared with others.

摘要

•Moment matching and graph embedding are leveraged in a unified framework for subspace learning.•Clustering and structural risk minimization (SRM) are jointly optimized to update soft labels for the target domain.•Extensive experiments on both multi-source and single-source domain adaptation show the superiority of our method compared with others.

论文关键词：Multi-source domain adaptation,Graph embedding,Subspace learning

论文评审过程：Received 9 May 2020, Revised 9 July 2020, Accepted 30 July 2020, Available online 18 August 2020, Version of Record 20 October 2020.

论文官网地址：https://doi.org/10.1016/j.ipm.2020.102367