Learning Dynamic Preference Structure Embedding From Temporal Networks

TL;DR

This paper introduces a novel framework called Dynamic Preference Structure (DPS) that captures evolving node preferences in temporal networks using adaptive sampling and graph fusion, improving downstream task performance.

Contribution

The paper proposes a new DPS framework with adaptive sampling and attention-based graph fusion to model dynamic node preferences in temporal networks.

Findings

DPS outperforms state-of-the-art methods on real-world temporal networks.

The adaptive sampling effectively captures dynamic preference structures.

Graph fusion enhances the quality of temporal node embeddings.

Abstract

The dynamics of temporal networks lie in the continuous interactions between nodes, which exhibit the dynamic node preferences with time elapsing. The challenges of mining temporal networks are thus two-fold: the dynamic structure of networks and the dynamic node preferences. In this paper, we investigate the dynamic graph sampling problem, aiming to capture the preference structure of nodes dynamically in cooperation with GNNs. Our proposed Dynamic Preference Structure (DPS) framework consists of two stages: structure sampling and graph fusion. In the first stage, two parameterized samplers are designed to learn the preference structure adaptively with network reconstruction tasks. In the second stage, an additional attention layer is designed to fuse two sampled temporal subgraphs of a node, generating temporal node embeddings for downstream tasks. Experimental results on many real-life temporal networks show that our DPS outperforms several state-of-the-art methods substantially owing to learning an adaptive preference structure. The code will be released soon at https://github.com/doujiang-zheng/Dynamic-Preference-Structure.