Online deep transferable dictionary learning

作者：

Highlights：

• We propose a two-level affiliation regularizer to reveal the local instance-level and global cluster-level affiliations.

• We propose an online deep transferable dictionary learning method to harness knowledge from incoming unlabeled data and previous labeled data.

• We verify the feasibility of our ODTDL approach in both online SSL and UDA scenarios.

摘要

•We propose a two-level affiliation regularizer to reveal the local instance-level and global cluster-level affiliations.•We propose an online deep transferable dictionary learning method to harness knowledge from incoming unlabeled data and previous labeled data.•We verify the feasibility of our ODTDL approach in both online SSL and UDA scenarios.

论文关键词：Online transferable dictionary learning,Semi-supervised learning,Domain adaptation

论文评审过程：Received 28 October 2020, Revised 1 April 2021, Accepted 27 April 2021, Available online 24 May 2021, Version of Record 4 June 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108007