Inductive learning utilises only the visual features and semantic information of the seen classes to build a model. In transductive setting, in addition to the seen class information, the semantic representations and unlabeled visual features of the unseen classes are exploited for learning.
————论文《A Review of Generalized Zero-Shot Learning
Methods》
版权声明：本文为qq_43197644原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。