Data Augmentation for Graph Neural Networks

Authors: Tong Zhao, Yozen Liu, Leonardo Neves, Oliver Woodford, Meng Jiang, Neil Shah11015-11023

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on multiple benchmarks show that augmentation via GAUG improves performance across GNN architectures and datasets.
Researcher Affiliation Collaboration Tong Zhao1*, Yozen Liu2, Leonardo Neves2, Oliver Woodford2, Meng Jiang1, Neil Shah2 1 University of Notre Dame, Notre Dame, IN 46556 2 Snap Inc., Santa Monica, CA 90405
Pseudocode No The paper describes the GAUG framework and its components, but it does not include structured pseudocode or an algorithm block.
Open Source Code Yes Our implementation is made publicly available at https://github.com/zhao-tong/GAug.
Open Datasets Yes We evaluate using 6 benchmark datasets across domains: citation networks (CORA, CITESEER (Kipf and Welling 2016a)), protein-protein interactions (PPI (Hamilton, Ying, and Leskovec 2017)), social networks (BLOGCATALOG, FLICKR (Huang, Li, and Hu 2017)), and air traffic (AIRUSA (Wu, He, and Xu 2019)).
Dataset Splits Yes We follow the semi-supervised setting in most GNN literature (Kipf and Welling 2016a; Veliˇckovi c et al. 2017) for train/validation/test splitting on CORA and CITESEER, and a 10/20/70% split on other datasets due to varying choices in prior work.
Hardware Specification No The paper does not specify the hardware (e.g., CPU, GPU models) used for running the experiments.
Software Dependencies No The paper mentions 'Optuna' for hyperparameter search but does not specify versions for any software dependencies used for the implementation or experiments.
Experiment Setup No The paper states 'employing Optuna (Akiba et al. 2019) for efficient hyperparameter search' but does not list the specific hyperparameter values or other detailed training configurations (e.g., learning rate, batch size, number of epochs) in the main text.