FedOGL framework combats catastrophic forgetting in federated graph learning

Researchers have developed FedOGL, a new framework designed to combat catastrophic forgetting in federated open-world multimodal graph learning. This approach aims to enable clients to learn new classes from private graph data while retaining historical knowledge and rejecting unknown samples. FedOGL employs replay and task-start distillation on the client side to preserve past decision behaviors and uses projection onto a shared structure basis to protect graph propagation memory. On the server side, it maintains compact category prototypes for cross-client knowledge sharing without exposing raw data. Experiments show FedOGL reduces performance degradation from catastrophic forgetting by 42.67% compared to existing methods. AI

IMPACT This framework could improve the robustness and efficiency of decentralized AI systems that handle evolving data.

RANK_REASON This is a research paper detailing a new framework for a specific machine learning problem. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• Connected Papers
• DagsHub
• FedOGL
• Gotit.pub
• Hugging Face
• IArxiv
• Litmaps
• ScienceCast
• scite Smart Citations

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →