Generalized Category Discovery Utilizing Reciprocal Learning and Class-wise Distribution Regularization

Duo Liu, Zhiquan Tan, Linglan Zhao, Zhongqiang Zhang, Xiangzhong Fang, Weiran Huang

Keywords: Generalized Category Discovery, Novel Class Discovery, Distribution Regularization

Abstract:

Generalized Category Discovery (GCD) aims to identify unlabeled samples by leveraging the base knowledge from labeled ones, where the unlabeled set consists of both base and novel classes.
Since clustering methods are time-consuming at inference, parametric-based approaches have become more popular.
However, recent parametric-based methods suffer inferior base discrimination due to the unreliable self-supervision.
To address this issue, we propose a Reciprocal Learning Framework (RLF) that introduces an auxiliary branch devoted to base classification.
During training, the main branch filters the pseudo-base samples to the auxiliary branch.
In response, the auxiliary branch provides more reliable soft labels for the main branch, leading to a virtuous cycle.
Furthermore, we introduce Class-wise Distribution Regularization (CDR) to mitigate the leaning bias towards base classes.
CDR essentially increases the prediction confidence of the unlabeled data and boosts the novel class performance.
Combined with both components, our method achieves superior performance in all classes with negligible extra computation.
Extensive experiments on seven GCD datasets validate the effectiveness of our method, e.g. delivering a notable 2.1\% improvement on the Stanford Cars dataset.
Our codes will be available upon acceptance.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 8893