Learning discriminative domain-invariant prototypes for generalized zero shot learning

作者：

Highlights：

• Learn Discriminative Domain-Invariant Prototypes to solve the domain shift problem.

• Orthogonal constraint makes all prototypes orthogonal to each other and dispersed.

• The learned prototypes distribute on the surface of a unit hyper-spherical.

• Experiments on four popular datasets show the effectiveness of the method.

摘要

•Learn Discriminative Domain-Invariant Prototypes to solve the domain shift problem.•Orthogonal constraint makes all prototypes orthogonal to each other and dispersed.•The learned prototypes distribute on the surface of a unit hyper-spherical.•Experiments on four popular datasets show the effectiveness of the method.

论文关键词：Generalized Zero Shot Learning (GZSL),Domain-invariant learning,Orthogonal constraint,Dictionary learning

论文评审过程：Received 3 December 2019, Revised 17 March 2020, Accepted 19 March 2020, Available online 24 March 2020, Version of Record 16 April 2020.

论文官网地址：https://doi.org/10.1016/j.knosys.2020.105796