Complex networks usually consist of dense-connected cliques, which are defined as communities. A community structure is a reflection of the local characteristics existing in the network topology, this makes community detection become an important research field to reveal the internal structural characteristics of networks. In this article, an information-based community detection approach MINC-NRL is proposed, which can be applied to both overlapping and non-overlapping community detection. MINC-NRL introduces network representation learning (NRL) to represent the target network as vectors, then generates a community evolution process based on these vectors to reduce the search space, and finally, finds the best community partition in this process using mutual information between network and communities (MINC). Experiments on real-world and synthetic data sets verifies the effectiveness of the approach in community detection, both on non-overlapping and overlapping tasks.

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.

Community structure is a network characteristic where nodes can be naturally divided into densely connected groups. Community structures are ubiquitous in social, biological, and technological networks. Revealing community structure in the network helps in the understanding of the topological associations and interactions of elements in the network, as well as helping to mine their potential information. However, this has been proven to be a difficult challenge. On the one hand, this is because there is no unified definition of the quality of a community; on the other hand, due to the complexity of the network, it is impossible to traverse all the possibilities of community partitions to find the best one. Aiming at performing high-accuracy community detection, an information-theoretic approach AMI-NRL was proposed. The approach first constructs a community evolution process based on the representation of the target network, then finds the most stable community structure during the evolution using an average-mutual-information-based criterion. The experiments show that the approach can effectively detect community structures on real-world datasets and synthetic datasets.

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.