Deep Contrastive Graph Learning with Clustering-Oriented Guidance

TL;DR

DCGL addresses clustering without a predefined graph by coupling a pseudo-siamese network with two clustering-oriented contrastive modules. It builds an initial graph and jointly learns structure and features through local and global graph branches, guided by feature- and centroid-focused losses to preserve discriminability and cluster structure. The method demonstrates superior performance on seven benchmarks against diverse baselines and shows robustness through ablations and parameter studies. This approach offers a scalable, unsupervised framework for graph-based clustering that mitigates initial-graph bias and directly enforces clustering relevance in learned representations.

Abstract

Graph Convolutional Network (GCN) has exhibited remarkable potential in improving graph-based clustering. To handle the general clustering scenario without a prior graph, these models estimate an initial graph beforehand to apply GCN. Throughout the literature, we have witnessed that 1) most models focus on the initial graph while neglecting the original features. Therefore, the discriminability of the learned representation may be corrupted by a low-quality initial graph; 2) the training procedure lacks effective clustering guidance, which may lead to the incorporation of clustering-irrelevant information into the learned graph. To tackle these problems, the Deep Contrastive Graph Learning (DCGL) model is proposed for general data clustering. Specifically, we establish a pseudo-siamese network, which incorporates auto-encoder with GCN to emphasize both the graph structure and the original features. On this basis, feature-level contrastive learning is introduced to enhance the discriminative capacity, and the relationship between samples and centroids is employed as the clustering-oriented guidance. Afterward, a two-branch graph learning mechanism is designed to extract the local and global structural relationships, which are further embedded into a unified graph under the cluster-level contrastive guidance. Experimental results on several benchmark datasets demonstrate the superiority of DCGL against state-of-the-art algorithms.

Figures (6)

• Figure 1: Overview of DCGL. Note that the decoder for $\mathbf{H}^{v_2}$ is omitted in this pipeline. First, we construct an initial graph to enable GCN. Second, we generate two latent representations by the pseudo-siamese network. Third, we derive two adjacency graphs from different perspectives, and then compute their cluster-level graph embeddings. Afterward, the framework is updated with the joint loss shown in Eq. (\ref{['eq:overall_loss']}). The final result is obtained by performing spectral clustering on the converged $\mathbf{S}^L$.
• Figure 2: Visualization of partial embeddings learned by DCGL on YaleB.
• Figure 3: Ablation results of feature-level contrastive learning on PIE.
• Figure 4: Ablation results of cluster-level contrastive learning on ORL.
• Figure 5: Ablation results of clustering guidance on four datasets.