The graph clustering algorithm for hard samples mining is a recent research hotspot. In the current algorithm, the main problems include the lack of a fusion mechanism for comparing methods and a sample pair weighting strategy. The algorithms ignore "false negative" samples within the view when sampling positive samples and disregards the utilization of graph-level information for clustering. To address the issue above, this paper proposed a graph clustering algorithm based on hard sample sampling joint contrast augmentation. Initially, utilizing an autoencoder to learn embeddings and designing a self-weighted contrast loss for representation learning. This loss unifies the strategies of node comparison and hard sample pair weighting across different views by utilizing the calculated pseudo-label, similarity, and confidence information. By adjusting the weights of sample pairs in different confidence regions, the loss function drives the model to focus on different types of hard samples to learn discriminative features, improve the consistency of intra-cluster representation and the distinctiveness of inter-cluster representation, and enhance the ability to discriminate samples. Additionally, the clustering network in the algorithm projects the graph-level representation to maximize the representation consistency of clusters under different views through cluster contrast loss. Finally, the two comparison losses combine and iteratively optimize through self-supervised training to complete clustering. In the comparison with 9 benchmark algorithms on 5 real datasets, this algorithm demonstrated superior performance on 4 authoritative indicators, highlighting its excellent clustering capabilities. Ablation experiments demonstrate the effectiveness and transferability of the two contrasting modules.

[2024-01-16] Accepted Paper /doc/2023.10.0521/v1