Node-Time Conditional Prompt Learning In Dynamic Graphs

TL;DR

This paper introduces DYGPROMPT, a novel prompt learning framework for dynamic graphs that captures evolving node-time patterns, bridging the gap between pre-training and downstream tasks like node classification.

Contribution

It proposes dual prompts and dual condition-nets to better model temporal dynamics and task objectives in dynamic graph neural networks.

Findings

Outperforms existing methods on four public datasets.

Effectively captures evolving node-time interactions.

Bridges the gap between pre-training and downstream tasks.

Abstract

Dynamic graphs capture evolving interactions between entities, such as in social networks, online learning platforms, and crowdsourcing projects. For dynamic graph modeling, dynamic graph neural networks (DGNNs) have emerged as a mainstream technique. However, they are generally pre-trained on the link prediction task, leaving a significant gap from the objectives of downstream tasks such as node classification. To bridge the gap, prompt-based learning has gained traction on graphs, but most existing efforts focus on static graphs, neglecting the evolution of dynamic graphs. In this paper, we propose DYGPROMPT, a novel pre-training and prompt learning framework for dynamic graph modeling. First, we design dual prompts to address the gap in both task objectives and temporal variations across pre-training and downstream tasks. Second, we recognize that node and time features mutually characterize each other, and propose dual condition-nets to model the evolving node-time patterns in downstream tasks. Finally, we thoroughly evaluate and analyze DYGPROMPT through extensive experiments on four public datasets.