Prior-Constrained Association Learning for Fine-Grained Generalized Category Discovery

TL;DR

This paper introduces a novel prior-constrained association learning approach for generalized category discovery, leveraging labeled data as a prior to improve clustering and representation learning of both known and unknown categories.

Contribution

It fully integrates prior knowledge into the association process, enhancing clustering accuracy and representation learning in GCD tasks, outperforming existing methods.

Findings

Significantly outperforms state-of-the-art methods on multiple benchmarks.

Effectively utilizes prior knowledge to guide association and clustering.

Improves representation learning through non-parametric prototypical contrast.

Abstract

This paper addresses generalized category discovery (GCD), the task of clustering unlabeled data from potentially known or unknown categories with the help of labeled instances from each known category. Compared to traditional semi-supervised learning, GCD is more challenging because unlabeled data could be from novel categories not appearing in labeled data. Current state-of-the-art methods typically learn a parametric classifier assisted by self-distillation. While being effective, these methods do not make use of cross-instance similarity to discover class-specific semantics which are essential for representation learning and category discovery. In this paper, we revisit the association-based paradigm and propose a Prior-constrained Association Learning method to capture and learn the semantic relations within data. In particular, the labeled data from known categories provides a unique prior for the association of unlabeled data. Unlike previous methods that only adopts the prior as a pre or post-clustering refinement, we fully incorporate the prior into the association process, and let it constrain the association towards a reliable grouping outcome. The estimated semantic groups are utilized through non-parametric prototypical contrast to enhance the representation learning. A further combination of both parametric and non-parametric classification complements each other and leads to a model that outperforms existing methods by a significant margin. On multiple GCD benchmarks, we perform extensive experiments and validate the effectiveness of our proposed method.