The current research on ensemble clustering mainly focuses on integration strategies, but the attention regarding the measurement and optimization of basic cluster is less emphasized. Based on the information entropy theory, this paper proposes a quality metric of basic cluster, and the clusterings are further selected by incorporating two-branch decisions and three-way decisions respectively. Determined by preset threshold(s), mechanism of two-branch based basic clustering filtering (BCF2BD) and three-way based basic clustering filtering (BCF3WD) are developed. Concretely, the basic clustering in BCF2BD is deleted if the quality metric of it is less than the preset threshold ξ, and the new clustering member is added to maintain the basic cluster set count. The basic clustering in BCF3WD is deleted if the quality metric of it is less than the preset threshold β, retained if the quality metric of it is greater than the preset threshold α, recalculated if the quality metric of it is greater than β and less than α. Both mechanism executed repeatedly until either non-decrement of basic clusters occurred or maximum iteration count reached. Contrastive experiments show that both methods of filtering algorithms can effectively improve the performance of ensemble clustering, and the three-way decisions filtering algorithm get less time consumption.

Deep graph clustering aims to discover community structures in attributed graphs without labels and is useful for downstream applications such as citation analysis. However, existing methods often cannot make full use of both node features and graph structures, especially when the structure contains noisy links or when attributes and topology are misaligned. Our objective is to learn a robust consensus embedding for attributed graphs under such attribute–topology view misalignment in a fully self-supervised manner. To this end, we propose GDGCA (Gated Deep Graph Clustering with Contrastive Alignment), which combines a dual-stream encoder with layer-wise gated fusion and cross-view contrastive alignment, followed by DEC-style self-training to refine cluster assignments. Experiments on multiple benchmark datasets show that GDGCA achieves competitive clustering performance compared with strong baselines. We further observe stable convergence under noisy or misaligned conditions, indicating improved robustness. Overall, GDGCA provides an effective self-supervised framework for reliable deep graph clustering on real-world attributed graphs.

Knowledge Graphs (KGs) are extensively used in recommendation systems and information retrieval but often suffer from incompleteness. A popular solution to this problem is multi-hop inference through a reinforcement learning framework, which provides an interpretable path for predicting missing links in KGs. Most previous work focuses on improving the performance of multi-hop link prediction. However, it has been observed that many multi-hop paths generated by these methods are irrational; they often fail to reasonably explain the predicted answer entities. To address this challenge, we introduce the Joint Multi-hop Link Prediction (JMLP) framework. The framework consists of a relation attention network and an entity attention network, which collaboratively generate the reasoning paths. The relation attention module utilizes an induction network to encode historical paths and employs the graph self-attention mechanism to refine the interaction of relation contextual information. The entity attention module uses the graph attention mechanism to obtain the aggregated contextual features and leverages self-attention to strengthen the correlation between local and global contextual entity features. Extensive experiments on five datasets validate the effectiveness of our approach, demonstrating significant improvements both in predictive performance and interpretability compared to state-of-the-art methods.

Graph clustering is an efficient method for deep clustering that utilizes graph convolution. Graph convolution effectively combines structure and content information, and lots of recent graph convolution-based methods have shown promising results in clustering performance on actual attribution networks. However, the established methods mainly employ a fixed graph convolution order, and limited studies have focused on the flexible choice of k-order graph convolution. When utilizing graph convolution with a fixed low order, only considering a few hops per node or neighbors within a set range of hops fails to maximize node relationships or account for the variations within the graphs. In this paper, we propose an adaptive method for graph clustering using a three-way decision idea. Our method enables the adaptive selection of k-order graph convolution for different graphs by searching for the k-order convolution kernel that best suits the subsequent clustering task. Additionally, our approach uses higher-order graph convolution to capture the global clustering structure. We assess the effectiveness of our approach through theoretical analysis and extensive experiments on benchmark datasets. Empirical evidence indicates that our method surpasses state-of-the-art techniques.