Abstract:Generalized Category Discovery (GCD) aims to classify both base and novel images using labeled base data. However, current approaches inadequately address the intrinsic optimization of the co-occurrence matrix $\bar{A}$ based on cosine similarity, failing to achieve zero base-novel regions and adequate sparsity in base and novel domains. To address these deficiencies, we propose a Non-Negative Generalized Category Discovery (NN-GCD) framework. It employs Symmetric Non-negative Matrix Factorization (SNMF) as a mathematical medium to prove the equivalence of optimal K-means with optimal SNMF, and the equivalence of SNMF solver with non-negative contrastive learning (NCL) optimization. Utilizing these theoretical equivalences, it reframes the optimization of $\bar{A}$ and K-means clustering as an NCL optimization problem. Moreover, to satisfy the non-negative constraints and make a GCD model converge to a near-optimal region, we propose a GELU activation function and an NMF NCE loss. To transition $\bar{A}$ from a suboptimal state to the desired $\bar{A}^*$, we introduce a hybrid sparse regularization approach to impose sparsity constraints. Experimental results show NN-GCD outperforms state-of-the-art methods on GCD benchmarks, achieving an average accuracy of 66.1\% on the Semantic Shift Benchmark, surpassing prior counterparts by 4.7\%.

What problem does this paper attempt to address?

This paper aims to solve several key problems in Generalized Category Discovery (GCD). Specifically: 1. **Optimizing the co - occurrence matrix**: Existing GCD methods have deficiencies in optimizing the co - occurrence matrix $\bar{A}$. In particular, the optimization method based on cosine similarity fails to achieve the zero - base - class - new - class area and sufficient sparsity. This leads to confusion between base classes and new classes and affects the classification performance. 2. **Improving category separation**: Current methods have difficulty achieving clear separation at the category level when fine - tuning self - supervised pre - trained models through Contrastive Learning (CL). This makes the similarity matrix between base classes and new classes lack the necessary sparsity and non - confusion. 3. **Optimizing K - means clustering**: K - means clustering plays a central role in GCD tasks, but existing methods fail to fully optimize K - means clustering to achieve the optimal co - occurrence matrix $\bar{A}^*$. To address these challenges, the paper proposes the Non - Negative Generalized Category Discovery (NN - GCD) framework. The main contributions of this framework include: 1. **Theoretical basis**: The theoretical equivalence between K - means clustering, Symmetric Non - negative Matrix Factorization (SNMF), and Non - negative Contrastive Learning (NCL) is established. This theoretical basis provides a clear optimization path for GCD tasks. 2. **NCL optimization**: The non - negative activation neuron mechanism, NMF NCE loss function, and hybrid sparse regularization method are introduced. These innovations are optimized at the network architecture and loss function levels, enabling the GCD model to better converge to the global optimal solution and improve the separation between base classes and new classes. 3. **Experimental verification**: The effectiveness of the NN - GCD framework is verified through experimental evaluations on seven established GCD benchmark datasets. The experimental results show that the average accuracy of NN - GCD on the Semantic Shift Benchmark reaches 66.1%, which is 4.7% higher than that of existing methods. In conclusion, through theoretical analysis and experimental verification, this paper proposes a new GCD framework, which significantly improves the recognition ability of base classes and new classes.

A Fresh Look at Generalized Category Discovery through Non-negative Matrix Factorization

Learning to Distinguish Samples for Generalized Category Discovery

Generalized Category Discovery with Clustering Assignment Consistency

Generalized Categories Discovery for Long-tailed Recognition

ImbaGCD: Imbalanced Generalized Category Discovery

Multimodal Generalized Category Discovery

Learning Semi-supervised Gaussian Mixture Models for Generalized Category Discovery

Efficient Representation Learning for Generalized Category Discovery

Contrastive Mean-Shift Learning for Generalized Category Discovery

Solving the Catastrophic Forgetting Problem in Generalized Category Discovery

Feature Extractor Optimization for Discriminative Representations in Generalized Category Discovery

Active Generalized Category Discovery

Let’s Start Over: Retraining with Selective Samples for Generalized Category Discovery

Parametric Classification for Generalized Category Discovery: A Baseline Study

Memory Consistency Guided Divide-and-Conquer Learning for Generalized Category Discovery

Federated Generalized Category Discovery

Contextuality Helps Representation Learning for Generalized Category Discovery

MetaGCD: Learning to Continually Learn in Generalized Category Discovery

Composing Novel Classes: A Concept-Driven Approach to Generalized Category Discovery

No Representation Rules Them All in Category Discovery

Prediction Consistency Regularization for Generalized Category Discovery