Abstract:Multi-granularity attributed network representation learning constructs a multi-granularity attributed network to extract multi-granularity features of an attributed network while preserving network’s structure and attributes’ informance. Note that the construction of multi-granularity attributed network affects the performance of multi-granularity attributed network representation learning applied to the node classification task. The existing methods of constructing multi-granularity attributed network ignore the information of overlapping nodes and weaken node classification ability. To solve this problem, we make full use of the overlapping nodes’ information to construct a multi-granularity attributed network, use the least information loss to define the optimal granularity, and propose a multi-granularity attributed network representation learning method (MANOC) for preserving overlapping community structure. Specifically, our method can quickly construct attributed networks of different granularities through overlapping nodes to learn node representations. That is, in the coarsening module, an overlapping community detection algorithm is adopted to find overlapping nodes and then we increase the weights of the edges formed by overlapping nodes. Furthermore, based on the similarity between nodes and communities, the attribute values of overlapping nodes are reasonably allocated, and information entropy and cross entropy are integrated to select the optimal granularity. Finally, we compare the proposed method with six representative network representation learning methods in achieving node classification tasks on five real network datasets. The experiments reveal that our method improves the average node classification accuracy by 30.67%, 24.05%, 11.19%, 18.80%, 21.55% and 0.46% compared to DeepWalk, Node2Vec, CAN, MILE, GraphZoom and HANE, respectively. In addition, we also demonstrate that node classification accuracies reach the maximum on the networks with optimal granularities.

Multi-granularity network representation learning on overlapping communities

Discovering the Representation Bottleneck of Graph Neural Networks from Multi-order Interactions

Network Representation Learning with Community Awareness and Its Applications in Brain Networks

Discovering and Explaining the Representation Bottleneck of Graph Neural Networks from Multi-order Interactions

Hierarchical attention network for attributed community detection of joint representation

NMLPA: Uncovering Overlapping Communities in Attributed Networks Via a Multi-Label Propagation Approach

Attributed Multi-Order Graph Convolutional Network for heterogeneous graphs

Network representation learning based on community-aware and adaptive random walk for overlapping community detection

Hierarchical community structure preserving approach for network embedding

Discovering the Representation Bottleneck of Graph Neural Networks

Attribute community detection based on latent representation learning and graph regularized non-negative matrix factorization

Imbalanced Graph Classification with Multi-scale Oversampling Graph Neural Networks

Learning with multi-resolution overlapping communities

Multi-graph aggregated graph neural network for heterogeneous graph representation learning

Multiple heterogeneous network representation learning based on multi-granularity fusion

Graph Convolutional Networks Meet Markov Random Fields: Semi-Supervised Community Detection in Attribute Networks

HybridGNN: Learning Hybrid Representation in Multiplex Heterogeneous Networks

Multi-label graph node classification with label attentive neighborhood convolution

Non-negative matrix factorization for overlapping community detection in directed weighted networks with sparse constraints.

ANGraph: Attribute-Interactive Neighborhood-Aggregative Graph Representation Learning

Discovering overlapping communities in multi-layer directed networks