Multiview latent space learning with progressively fine-tuned deep features for unsupervised domain adaptation

Chenyang Zhu, Qian Wang, Yunxin Xie, Shoukun Xu

Abstract: Highlights•Propose using three views from a single model to extract features for domain alignment support in UDA and MDA scenarios.•A multiview latent space learning approach that leverages the constructed three views to enhance UDA and MDA tasks.•Select highly confident labeled target samples based on maximum conditional probability for pseudo-labeling.