Heterogeneous domain adaptation via incremental discriminative knowledge consistency

Yuwu Lu, Dewei Lin, Jiajun Wen, Linlin Shen, Xuelong Li, Zhenkun Wen

Abstract: Highlights•Distribution alighment is learnt to preserve discriminative knowledge consistency.•The data attribute can be preserved by optimizing the reconstruction loss.•Pseudo labeling strategy is developed for classifying unlabeled target samples.•The developed IDKC framework outperforms the state-of-the-art methods.