R2LP

This is the official repository of paper "Resurrecting Label Propagation for Graphs with Heterophily and Label Noise".

Setup

• torch==1.12.1
• networkx==2.6.3
• scipy==1.7.3
• numpy==1.21.5

Repository structure

├── data  # old datasets, including cora, citeseer, pubmed
├── new_data # new datasets, including texas, wisconsin, cornell, actor and chameleon
├── splits # Split part of dataset
├── train.py # the main code
├── model.py # model implementations
├── utils.py # preprocessing subroutines
├── run.sh # run results

Run pipeline

1. You can run R2LP by the script:

python R2LP/train.py --dataset cora

2. For more experiments running details, you can ref the running .sh by the script:

sh R2LP/run.sh