CTRL: Continuous-Time Representation Learning on Temporal Heterogeneous Information Network

TL;DR

CTRL tackles inductive continuous-time representation learning on temporal HINs by integrating three mechanisms within each layer: heterogeneous attention for semantic alignment, an edge-based Hawkes process to capture time-decayed influence across heterogeneous edges, and dynamic centrality to reflect evolving node importance. The model is trained with a future-event prediction objective that jointly models high-order subgraph events and constituent edges, enabling faithful evolution of network structure in continuous time. Experiments on ACM, DBLP, and IMDB demonstrate that CTRL consistently outperforms state-of-the-art baselines in inductive temporal link prediction, with ablation studies confirming the contributions of event-based training, dynamic centrality, and Hawkes-based temporal dynamics. This work provides a scalable, inductive framework for learning time-aware representations in complex, heterogeneous networks, with potential applications in evolving scholarly and multimedia graphs.

Abstract

Inductive representation learning on temporal heterogeneous graphs is crucial for scalable deep learning on heterogeneous information networks (HINs) which are time-varying, such as citation networks. However, most existing approaches are not inductive and thus cannot handle new nodes or edges. Moreover, previous temporal graph embedding methods are often trained with the temporal link prediction task to simulate the link formation process of temporal graphs, while ignoring the evolution of high-order topological structures on temporal graphs. To fill these gaps, we propose a Continuous-Time Representation Learning (CTRL) model on temporal HINs. To preserve heterogeneous node features and temporal structures, CTRL integrates three parts in a single layer, they are 1) a \emph{heterogeneous attention} unit that measures the semantic correlation between nodes, 2) a \emph{edge-based Hawkes process} to capture temporal influence between heterogeneous nodes, and 3) \emph{dynamic centrality} that indicates the dynamic importance of a node. We train the CTRL model with a future event (a subgraph) prediction task to capture the evolution of the high-order network structure. Extensive experiments have been conducted on three benchmark datasets. The results demonstrate that our model significantly boosts performance and outperforms various state-of-the-art approaches. Ablation studies are conducted to demonstrate the effectiveness of the model design.

Figures (5)

• Figure 1: The architecture of a layer of the CTRL model.
• Figure 2: AUC of the proposed method and temporal network embedding baselines on the three datasets.
• Figure 3: AUC scores of ablation studies.
• Figure 4: Dataset details of the ACM dataset.
• Figure 5: Dataset details of the DBLP dataset.