Posterior Label Smoothing for Node Classification

This repository is the official implementation of "Posterior Label Smoothing for Node Classification" accepted by AAAI 2026.

Abstract

Label smoothing is a widely studied regularization technique in machine learning. However, its potential for node classification in graph-structured data, spanning homophilic to heterophilic graphs, remains largely unexplored. We introduce posterior label smoothing, a novel method for transductive node classification that derives soft labels from a posterior distribution conditioned on neighborhood labels. The likelihood and prior distributions are estimated from the global statistics of the graph structure, allowing our approach to adapt naturally to various graph properties. We evaluate our method on 10 benchmark datasets using eight baseline models, demonstrating consistent improvements in classification accuracy. The following analysis demonstrates that soft labels mitigate overfitting during training, leading to better generalization performance, and that pseudo-labeling effectively refines the global label statistics of the graph.

Environment Setup

To set up the environment, use the configuration specified in the environment.yml file. Run the following command:

conda env create --file environment.yml 

Training GNNs with PosteL

To train GNNs using the PosteL, use the following command:

python training.py --labeling_method=postel --soft_label_ratio=$ALPHA --smoothing_ratio=$BETA --dataset=$DATASET --net=$GNN

Citation

Please cite our paper if you use the model or this code in your own work:

@article{heo2024posterior,
  title={Posterior Label Smoothing for Node Classification},
  author={Heo, Jaeseung and Park, Moonjeong and Kim, Dongwoo},
  journal={arXiv preprint arXiv:2406.00410},
  year={2024}
}