Reproduce “SUCRe: Selective Uncertainty-Aware Contrastive Representation for Graph Transfer 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
Mingcan Wang, Junchang Xin, Zhongming Yao, Bing Tian Dai, Kaifu Long, Zhiqiong Wang
The paper presents SUCRe, a graph transfer method for adapting a graph neural network from a labeled source graph to a sparsely labeled target graph. It estimates local ambiguity during feature alignment and selects semi-hard negative examples for contrastive learning. On 20-shot CiteSeer after Cora pretraining, it reaches 0.7332 accuracy versus 0.6928 for GraphLoRA, an improvement limited to this setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.