ProtoGCD: Unified and Unbiased Prototype Learning for Generalized Category Discovery
This addresses the challenge of imbalanced accuracy and biased representations in GCD, which is important for practical applications in machine learning, though it appears incremental as it builds on prior methods like pseudo-labeling and contrastive learning.
The paper tackles the problem of generalized category discovery (GCD), where models must cluster and discover novel categories using labeled samples from old classes, by introducing ProtoGCD, a unified prototype learning framework that achieves state-of-the-art performance on generic and fine-grained datasets.
Generalized category discovery (GCD) is a pragmatic but underexplored problem, which requires models to automatically cluster and discover novel categories by leveraging the labeled samples from old classes. The challenge is that unlabeled data contain both old and new classes. Early works leveraging pseudo-labeling with parametric classifiers handle old and new classes separately, which brings about imbalanced accuracy between them. Recent methods employing contrastive learning neglect potential positives and are decoupled from the clustering objective, leading to biased representations and sub-optimal results. To address these issues, we introduce a unified and unbiased prototype learning framework, namely ProtoGCD, wherein old and new classes are modeled with joint prototypes and unified learning objectives, {enabling unified modeling between old and new classes}. Specifically, we propose a dual-level adaptive pseudo-labeling mechanism to mitigate confirmation bias, together with two regularization terms to collectively help learn more suitable representations for GCD. Moreover, for practical considerations, we devise a criterion to estimate the number of new classes. Furthermore, we extend ProtoGCD to detect unseen outliers, achieving task-level unification. Comprehensive experiments show that ProtoGCD achieves state-of-the-art performance on both generic and fine-grained datasets. The code is available at https://github.com/mashijie1028/ProtoGCD.