Unleashing the Power of Neural Collapse: Consistent Supervised-Unsupervised Alignment for Generalized Category Discovery

2025-07-07

AI Summary

Generalized Category Discovery (GCD) suffers from feature confusion and class overlap due to misalignment between supervised and unsupervised learning objectives, severely hindering novel-class recognition. To address this, we propose a neural collapse-inspired geometric structure optimization framework. It employs Equiangular Tight Frame (ETF) prototypes to construct a unified feature space, enabling consistent alignment of both known and unknown classes. We further introduce a supervised-unsupervised joint alignment loss and a semantic consistency matcher to enhance inter-class separability and intra-class clustering stability. Our method requires no additional annotations or strong assumptions. Extensive experiments on multiple GCD benchmarks demonstrate significant improvements in novel-class accuracy—e.g., +3.2% on CIFAR-100 and +4.7% on ImageNet-100—validating its effectiveness and robustness across diverse settings.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Learning & Optimization for CV
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

Generalized Category Discovery (GCD) focuses on classifying known categories while simultaneously discovering novel categories from unlabeled data. However, previous GCD methods face challenges due to inconsistent optimization objectives and category confusion. This leads to feature overlap and ultimately hinders performance on novel categories. To address these issues, we propose the Neural Collapse-inspired Generalized Category Discovery (NC-GCD) framework. By pre-assigning and fixing Equiangular Tight Frame (ETF) prototypes, our method ensures an optimal geometric structure and a consistent optimization objective for both known and novel categories. We introduce a Consistent ETF Alignment Loss that unifies supervised and unsupervised ETF alignment and enhances category separability. Additionally, a Semantic Consistency Matcher (SCM) is designed to maintain stable and consistent label assignments across clustering iterations. Our method achieves strong performance on multiple GCD benchmarks, significantly enhancing novel category accuracy and demonstrating its effectiveness.

Problem

• Classify known and discover novel categories from unlabeled data
• Address inconsistent optimization objectives and category confusion
• Enhance category separability and stable label assignments

Innovation

• Pre-assigned ETF prototypes ensure optimal geometric structure
• Consistent ETF Alignment Loss unifies supervised and unsupervised learning
• Semantic Consistency Matcher stabilizes label assignments across iterations