Restructuring Graph for Higher Homophily via Adaptive Spectral Clustering

Authors: Shouheng Li, Dongwoo Kim, Qing Wang

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The experimental results show that our graph restructuring method can significantly boost the performance of six classical GNNs by an average of 25% on less-homophilic graphs.
Researcher Affiliation Academia 1 School of Computing, Australian National University, Canberra, Australia 2 CSE & GSAI, POSTECH, Pohang, South Korea 3 Data61, CSIRO, Canberra, Australia
Pseudocode Yes A detailed algorithm can be found in Appendix.
Open Source Code Yes The code is available at https://github.com/seanli3/graph_restructure.
Open Datasets Yes We evaluate on four real-world graphs: TEXAS, CORNELL, CHAMELEON and SQUIRREL (Rozemberczki, Allen, and Sarkar 2021), as well as synthetic graphs of controlled homophily.
Dataset Splits Yes We adopt early stopping and record the results from the epoch with highest validation accuracy. We use the same split setting as Pei et al. (2020); Zhu et al. (2020).
Hardware Specification Yes All experiments are run on a single NVIDIA RTX A6000 48GB GPU unless otherwise noted.
Software Dependencies No The paper does not provide specific software dependency versions (e.g., library or framework names with version numbers).
Experiment Setup Yes Hyperparameters are tuned using grid search for all models on the unmodified and restructured graphs of each dataset. For the spectrum slicer in Equation 10, we use a set of 20 slicers with s = 40 and m = 4 so that the spectrum is sliced into 20 even range of 0.1.