ICML 2019oral104 citations

Co-Representation Network for Generalized Zero-Shot Learning

Fei Zhang, Guangming Shi

Abstract

Generalized zero-shot learning is a significant topic but faced with bias problem, which leads to unseen classes being easily misclassified into seen classes. Hence we propose a embedding model called co-representation network to learn a more uniform visual embedding space that effectively alleviates the bias problem and helps with classification. We mathematically analyze our model and find it learns a projection with high local linearity, which is proved to cause less bias problem. The network consists of a cooperation module for representation and a relation module for classification, it is simple in structure and can be easily trained in an end-to-end manner. Experiments show that our method outperforms existing generalized zero-shot learning methods on several benchmark datasets.

BibTeX
@InProceedings{pmlr-v97-zhang19l,
  title = 	 {Co-Representation Network for Generalized Zero-Shot Learning},
  author =       {Zhang, Fei and Shi, Guangming},
  booktitle = 	 {Proceedings of the 36th International Conference on Machine Learning},
  pages = 	 {7434--7443},
  year = 	 {2019},
  editor = 	 {Chaudhuri, Kamalika and Salakhutdinov, Ruslan},
  volume = 	 {97},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {09--15 Jun},
  publisher =    {PMLR},
  pdf = 	 {http://proceedings.mlr.press/v97/zhang19l/zhang19l.pdf},
  url = 	 {https://proceedings.mlr.press/v97/zhang19l.html},
  abstract = 	 {Generalized zero-shot learning is a significant topic but faced with bias problem, which leads to unseen classes being easily misclassified into seen classes. Hence we propose a embedding model called co-representation network to learn a more uniform visual embedding space that effectively alleviates the bias problem and helps with classification. We mathematically analyze our model and find it learns a projection with high local linearity, which is proved to cause less bias problem. The network consists of a cooperation module for representation and a relation module for classification, it is simple in structure and can be easily trained in an end-to-end manner. Experiments show that our method outperforms existing generalized zero-shot learning methods on several benchmark datasets.}
}