Reproduce “GPT-GNN: Generative Pre-Training of Graph Neural Networks” 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

Ziniu Hu, Yuxiao Dong, Kuansan Wang, Kai-Wei Chang, Yizhou Sun

GPT-GNN pre-trains graph neural networks on unlabeled graphs by reconstructing node attributes and links, then fine-tunes them for a task. It uses neighboring links to infer missing attributes, then those attributes to infer masked links. On OAG Paper–Field prediction under combined time-and-field transfer, HGT with 10% labels achieves MRR 0.393 ± 0.112, versus 0.336 ± 0.149 without pre-training.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.