Publication
Fast graph condensation with structure-based neural tangent kernel
Graph Condensation
Cite this work View paper
Citation
Lin Wang, Wenqi Fan, Jiatong Li, Yao Ma, and Qing Li. (2023). Fast graph condensation with structure-based neural tangent kernel. arXiv preprint arXiv:2310.11046. (Accepted by WWW 24)
About this work
GC-SNTK formulates graph condensation as kernel ridge regression. A structure-based neural tangent kernel captures graph topology and removes the need to repeatedly train a graph neural network in the inner optimization loop.
The experiments measure condensation speed and downstream prediction performance on the condensed graphs.