High-Order Topology-Enhanced Graph Convolutional Networks for Dynamic Graphs

Abstract

Understanding the evolutionary mechanisms of dynamic graphs is crucial since dynamic is a basic characteristic of real-world networks. The challenges of modeling dynamic graphs are as follows: (1) Real-world dynamics are frequently characterized by group effects, which essentially emerge from high-order interactions involving groups of entities. Therefore, the pairwise interactions revealed by the edges of graphs are insufficient to describe complex systems. (2) The graph data obtained from real systems are often noisy, and the spurious edges can interfere with the stability and efficiency of models. To address these issues, we propose a high-order topology-enhanced graph convolutional network for modeling dynamic graphs. The rationale behind it is that the symmetric substructure in a graph, called the maximal clique, can reflect group impacts from high-order interactions on the one hand, while not being readily disturbed by spurious links on the other hand. Then, we utilize two independent branches to model the distinct influence mechanisms of the two effects. Learnable parameters are used to tune the relative importance of the two effects during the process. We conduct link predictions on real-world datasets, including one social network and two citation networks. Results show that the average improvements of the high-order enhanced methods are 68%, 15%, and 280% over the corresponding backbones across datasets. The ablation study and perturbation analysis validate the effectiveness and robustness of the proposed method. Our research reveals that high-order structures provide new perspectives for studying the dynamics of graphs and highlight the necessity of employing higher-order topologies in the future.

Heterogeneous network approaches to protein pathway prediction

“…Magnano et al conducted a preliminary case study in human fibroblasts to predict localized interactions involved in the pathway that responds to viral infection [94] . This type of dynamic data has been used in social networks to train a DGCN for link prediction, and these methods are applicable to the PPI network as well [95] .…”

Motif Prediction with Graph Neural Networks

“…There have been efforts into combining HO with temporal graph data models [93]. This includes graph data models such as d-dimensional De Bruijn graphs [156], [215], memory networks [207], HO Markov models for temporal networks [24], [194], [253], motif-based process representations [216], multi-layer and multiplex networks [151], hypergraphs [75], [245], [260], [266], and others [286].…”

A Novel Money Laundering Prediction Model Based on a Dynamic Graph Convolutional Neural Network and Long Short-Term Memory

“…Nevertheless, the current research relies primarily on the static design of money laundering transaction graphs and uses static graphs for graph convolution purposes in dynamic time series prediction tasks, leading to issues such as dynamic temporal feature neglect and information leakage. Recent studies [37,38] have proposed an innovative approach in this context, introducing a time attribute based on static graphs and transforming them into dynamic graphs based on timestamps to enhance the predictive performance of the constructed model and mitigate the potential risk of excessive smoothing. The neural network architecture constructed based on the time-series dynamic model consistently exhibits symmetrical properties at each time step, ensuring persistent symmetry in the model, which in turn guarantees its robustness and credibility.…”