Dissecting the Failure of Invariant Learning on Graphs

Authors: Qixun Wang, Yifei Wang, Yisen Wang, Xianghua Ying

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on graph OOD benchmarks validate the superiority of CIA and CIA-LRA, marking a significant advancement in nodelevel OOD generalization. ... 5 Experiments
Researcher Affiliation Academia Qixun Wang1 Yifei Wang 2 Yisen Wang 1,3 Xianghua Ying 1 1 State Key Laboratory of General Artificial Intelligence, School of Intelligence Science and Technology, Peking University 2 CSAIL, MIT 3 Institute for Artificial Intelligence, Peking University
Pseudocode Yes Algorithm 1 Detailed Training Procedure of CIA-LRA
Open Source Code Yes The codes are available at https://github.com/NOVAglow646/Neur IPS24-Invariant-Learning-on-Graphs.
Open Datasets Yes We run experiments using 3-layer GAT and GCN on GOOD [Gui et al., 2022], a graph OOD benchmark.
Dataset Splits Yes Following Gui et al. [2022], we use an OOD validation set for model selection.
Hardware Specification Yes We use one NVIDIA Ge Force RTX 3090 or 4090 GPU for each single experiments.
Software Dependencies No The paper mentions using GAT and GCN models but does not specify versions for software dependencies like PyTorch, CUDA, or Python libraries.
Experiment Setup Yes The detailed experimental setup and hyperparameter settings are in Appendix C.