Node-Time Conditional Prompt Learning In Dynamic Graphs

2024-05-22

AI Summary

To address the dual mismatch—between pretraining objectives (e.g., link prediction) and downstream tasks (e.g., node classification) in both objective functions and temporal modeling—this paper proposes DYGPROMPT, a novel framework for dynamic graph neural network (DGNN) pretraining. Its core innovation is the first-ever node-time bidirectional conditional prompting mechanism: two dedicated prompt modules explicitly model node semantics and temporal evolution, while a conditional generation network jointly aligns time-aware representations across pretraining and downstream tasks, effectively bridging objective discrepancy and temporal misalignment. Extensive experiments on four standard dynamic graph benchmarks demonstrate that DYGPROMPT significantly improves downstream node classification performance, validating the effectiveness, robustness, and generalizability of its time-aware prompting mechanism.

Technology Category

• Natural Language Processing: Prompt Engineering / Prompting
• Machine Learning: Graph-based Machine Learning
• Planning, Routing, and Scheduling: Temporal Planning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

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.

Problem

• Bridges gap between pre-training and downstream tasks in dynamic graphs.
• Addresses temporal variations and task objective mismatches in dynamic graph modeling.
• Models evolving node-time patterns using dual condition-nets and prompts.

Innovation

• Dual prompts bridge task and temporal gaps
• Dual condition-nets model node-time patterns
• DYGPROMPT framework for dynamic graph learning