More Domains, No Gains? FLO: Fast-Slow Reciprocal Learning for Multi-Domain Graph Pre-training

Key points

• Exhibiting such intelligence requires effectively learning knowledge from large-scale, multi-domain graph data during pre-training.1
• Most existing methods pursue this goal by designing domain alignment strategies that help a single graph model learn from multiple domains.1
• To address these issues, we propose FLO, a Fast--sLow reciprOcal approach for multi-domain graph pre-training.1
• We further stabilize knowledge acquisition by pairing a fast process that adapts to the current domain with a slow process that integrates knowledge across domains.1

Sources (1)

• [1]More Domains, No Gains? FLO: Fast-Slow Reciprocal Learning for Multi-Domain Graph Pre-training
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 10, 09:12 AM

  Inspired by the great success of foundation models in the development of artificial general intelligence, Graph Foundation Models (GFMs) have recently attracted significant attention and have the potential to advance graph intelligence.

  Exhibiting such intelligence requires effectively learning knowledge from large-scale, multi-domain graph data during pre-training.

Extractive summary: sentences quoted from the sources.