Structural–Temporal embedding of large-scale dynamic networks with parallel implementation. (May 2022)

• Record Type: Journal Article
• Authors: Xie, Luodie; Shen, Hong; Feng, Dawei
• Is Part Of: Computers & electrical engineering. Volume 100(2022)
• Publication Date: 2022-05
• Subjects: Temporal network embedding -- Structural similarity -- Temporal information -- Hawkes process -- Parallel training; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2022.107835
• Languages: English
• ISSNs: 0045-7906

Abstract

Abstract: Due to the widespread network data in the real world, network analysis has attracted increasing attention in recent years. In complex systems such as social networks, entities and their mutual relations can be respectively represented by nodes and edges composing a network. Because occurrences of entities and relations in these systems are often dynamic over time, their networks are called temporal networks describing the process of dynamic connection of nodes in the networks. Dynamic network embedding aims to embed nodes in a temporal network into a low-dimensional semantic space, such that the network structures and evolution patterns can be preserved as much as possible in the latent space. Most existing methods capture structural similarities (relations) of strongly-connected nodes based on their historical neighborhood information, they ignore the structural similarities of weakly-connected nodes that may also represent relations and include no explicit temporal information in node embeddings for capturing periodic dependency of events. To address these issues, we propose a novel temporal network embedding model by extending the structure similarity to cover both strong connections and weak connections among nodes, and including the temporal information in node embeddings. To improve the training efficiency of our model, we present a parallel training strategy to quickly acquire node embeddings. Extensive experiments on several real-world temporal networks demonstrate that our model significantly outperforms the state-of-the-arts in traditional tasks, including link prediction and node classification.