Reproduce “Prompt-Driven Continual Graph Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Qi Wang, Tianfei Zhou, Ye Yuan, Rui Mao
PROMPTCGL introduces a prompt-driven framework for Continual Graph Learning, achieving state-of-the-art performance with constant memory usage. This approach effectively mitigates catastrophic forgetting and preserves data privacy by learning task-specific prompts rather than replaying historical graph data.
Key insights
- Graph prompt learning provides an effective and memory-efficient alternative to memory replay for mitigating catastrophic forgetting in continual graph learning settings.
- Hierarchical prompting, addressing both node features and local topological structures, is crucial for enabling GNNs to adapt to diverse and evolving graph tasks.
- A personalized prompt generation mechanism allows for dynamic, context-aware instruction to the GNN backbone while maintaining constant memory complexity, crucial for scalability and privacy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.