Abstract:The rich information underlying graphs has inspired further investigation of unsupervised graph representation. Existing studies mainly depend on node features and topological properties within static graphs to create self-supervised signals, neglecting the temporal components carried by real-world graph data, such as timestamps of edges. To overcome this limitation, this paper explores how to model temporal evolution on dynamic graphs elegantly. Specifically, we introduce a new inductive bias, namely temporal translation invariance, which illustrates the tendency of the identical node to keep similar labels across different timespans. Based on this assumption, we develop a dynamic graph representation framework CLDG that encourages the node to maintain locally consistent temporal translation invariance through contrastive learning on different timespans. Except for standard CLDG which only considers explicit topological links, our further proposed CLDG＋＋additionally employs graph diffusion to uncover global contextual correlations between nodes, and designs a multi-scale contrastive learning objective composed of local-local, local-global, and global-global contrasts to enhance representation capabilities. Interestingly, by measuring the consistency between different timespans to shape anomaly indicators, CLDG and CLDG＋＋are seamlessly integrated with the task of spotting anomalies on dynamic graphs, which has broad applications in many high-impact domains, such as finance, cybersecurity, and healthcare. Experiments demonstrate that CLDG and CLDG＋＋both exhibit desirable performance in downstream tasks including node classification and dynamic graph anomaly detection. Moreover, CLDG significantly reduces time and space complexity by implicitly exploiting temporal cues instead of complicated sequence models. The code and data are available at https://github.com/yimingxu24/CLDG.

RDGSL: Dynamic Graph Representation Learning with Structure Learning

Dynamic Graph Representation Learning via Edge Temporal States Modeling and Structure-reinforced Transformer

Dynamic Representation Learning Via Recurrent Graph Neural Networks

Time-aware Graph Structure Learning Via Sequence Prediction on Temporal Graphs.

SiGNN: A Spike-induced Graph Neural Network for Dynamic Graph Representation Learning

DGSLN: Differentiable Graph Structure Learning Neural Network for Robust Graph Representations.

DG-Mamba: Robust and Efficient Dynamic Graph Structure Learning with Selective State Space Models

Expressivity of Representation Learning on Continuous-Time Dynamic Graphs: An Information-Flow Centric Review

Fuzzy Representation Learning on Dynamic Graphs

Learning Continuous Dynamic Network Representation with Transformer-Based Temporal Graph Neural Network.

Directed Acyclic Graph Structure Learning from Dynamic Graphs

Graph Structure Learning for Robust Recommendation

Learning dynamic graph representations through timespan view contrasts

DGSTD: Learning on Dynamic Graph with Spatio-Temporal Disentanglement

A Hybrid Continuous-Time Dynamic Graph Representation Learning Model by Exploring Both Temporal and Repetitive Information

Scalable and Effective Temporal Graph Representation Learning With Hyperbolic Geometry

SE-GSL: A General and Effective Graph Structure Learning Framework Through Structural Entropy Optimization

Spatial-temporal Graph Convolutional Networks with Diversified Transformation for Dynamic Graph Representation Learning

Structure-Preserving Graph Representation Learning

A recurrent graph neural network for inductive representation learning on dynamic graphs

CLDG: Contrastive Learning on Dynamic Graphs.