Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Rethinking Spectral Augmentation for Contrast-based Graph Self-Supervised Learning
Authors: Xiangru Jian, Xinjian Zhao, Wei Pang, Chaolong Ying, Yimu Wang, Yaoyao Xu, Tianshu Yu
TMLR 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Through extensive empirical studies, we find that simple edge perturbations random edge dropping for node-level and random edge adding for graph-level self-supervised learning consistently yield comparable or superior performance while being significantly more computationally efficient. This suggests that the computational overhead of sophisticated spectral augmentations may not justify their practical benefits. Our theoretical analysis of the Info NCE loss bounds for shallow GNNs further supports this observation. And In Sec 5, we claim that simple edge perturbation techniques, like adding edges to or dropping edges from the graph, not only compete well but often outperform spectral augmentations, without any significant help from spectral cues. To support this, (a) In Sec. 6, overall model performance on test accuracy with four state-of-the-art frameworks on both nodeand graph-level classification tasks support the superiority of simple edge perturbation. (b) Studies in Sec. 7.1 reveal the indistinguishability between the average spectrum of augmented graphs from edge perturbation with optimal parameters on different datasets, no matter how different that of original graphs is, indicating GNN encoders can hardly learn spectral information from augmented graphs. (c) In Sec. 7.2, we analyze the effectiveness of state-of-the-art spectral augmentation baseline (i.e., SPAN) by perturbing edges to alter the spectral characteristics of augmented graphs from simple edge perturbation augmentation and examining the impact on model performance. (d) In Appendix E.3, statistical analysis is carried out to argue that the major reason edge perturbation works well is not because of the spectral information as they are statistically not the key factor on model performance. |
| Researcher Affiliation | Academia | Xiangru Jian EMAIL Cheriton School of Computer Science University of Waterloo, Xinjian Zhao EMAIL School of Data Science The Chinese University of Hong Kong, Shenzhen, Wei Pang EMAIL Cheriton School of Computer Science University of Waterloo Vector Institute, Chaolong Ying EMAIL School of Data Science The Chinese University of Hong Kong, Shenzhen, Yimu Wang EMAIL Cheriton School of Computer Science University of Waterloo, Yaoyao Xu EMAIL School of Data Science The Chinese University of Hong Kong, Shenzhen, Tianshu Yu EMAIL School of Data Science The Chinese University of Hong Kong, Shenzhen |
| Pseudocode | No | The paper describes methods and theoretical analysis in paragraph form and through mathematical equations, but it does not contain any explicitly labeled 'Pseudocode', 'Algorithm', or code-like formatted procedure blocks. |
| Open Source Code | No | The paper does not provide an explicit statement about releasing its own source code for the methodology described, nor does it provide a direct link to a code repository for its implementation. It mentions using and implementing frameworks 'based on' an existing library (PyGCL) for experiments, but this is a third-party tool, not the authors' specific code for this research. |
| Open Datasets | Yes | We conducted extensive experiments for node-level classification on seven datasets: Cora, Cite Seer, Pub Med (Kipf & Welling, 2016), Photo, Computers (Shchur et al., 2018), Coauthor-CS, and Coauthor-Phy. These datasets include various types of graphs, such as citation networks, co-purchase networks, and co-authored networks. [...] Additionally, we carried out graph-level classification on five datasets from the TUDataset collection (Morris et al., 2020), which include biochemical molecules and social networks. [...] All datasets can be accessed through PyG library 1. 1https://pytorch-geometric.readthedocs.io/en/latest/modules/datasets.html |
| Dataset Splits | Yes | For the node classification task nodes are randomly divided into 10%/10%/80% for training, validation, and testing, and for graph classification datasets, graphs are randomly divided into 80%/10%/10% for training, validation, and testing. |
| Hardware Specification | Yes | All experiments are conducted using 8 NVIDIA A100 GPU. |
| Software Dependencies | No | The paper states, 'For each CG-SSL framework, we implement it based on (Zhu et al., 2021a) 2. 2https://github.com/PyGCL/PyGCL'. However, it does not provide specific version numbers for PyGCL or any other software libraries, programming languages, or environments used in the experiments. |
| Experiment Setup | Yes | We use the following hyperparameters: the learning rate is set to 5 10 4, and the node hidden size is set to 512, the number of GCN encoder layer is set {1, 2}. For all node classification datasets, training epochs are set {50, 100, 150, 200, 400, 1000}, and for all graph classification datasets, training epochs are set {20, 40, ..., 200}. To achieve performance closer to the global optimum, we use randomized search to determine the optimal probability of edge perturbation and SPAN perturbation ratio. For Cora and Cite Seer the search is conducted one hundred times, and for all other datasets, it is conducted twenty times. For all graph classification datasets, the batch size is set to 128. |