ReLKD: Inter-Class Relation Learning with Knowledge Distillation for Generalized Category Discovery
This work addresses a challenge in machine learning for real-world scenarios where data includes both known and novel classes, but it is incremental as it builds on existing GCD methods by incorporating inter-class relations.
The paper tackles the problem of Generalized Category Discovery (GCD), which involves categorizing unlabeled data with both known and novel classes using only labels for known classes, by proposing ReLKD, a framework that exploits implicit inter-class relations to enhance novel class classification, showing effectiveness in scenarios with limited labeled data across four datasets.
Generalized Category Discovery (GCD) faces the challenge of categorizing unlabeled data containing both known and novel classes, given only labels for known classes. Previous studies often treat each class independently, neglecting the inherent inter-class relations. Obtaining such inter-class relations directly presents a significant challenge in real-world scenarios. To address this issue, we propose ReLKD, an end-to-end framework that effectively exploits implicit inter-class relations and leverages this knowledge to enhance the classification of novel classes. ReLKD comprises three key modules: a target-grained module for learning discriminative representations, a coarse-grained module for capturing hierarchical class relations, and a distillation module for transferring knowledge from the coarse-grained module to refine the target-grained module's representation learning. Extensive experiments on four datasets demonstrate the effectiveness of ReLKD, particularly in scenarios with limited labeled data. The code for ReLKD is available at https://github.com/ZhouF-ECNU/ReLKD.