Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Structure-free Graph Condensation: From Large-scale Graphs to Condensed Graph-free Data

About

Graph condensation, which reduces the size of a large-scale graph by synthesizing a small-scale condensed graph as its substitution, has immediate benefits for various graph learning tasks. However, existing graph condensation methods rely on the joint optimization of nodes and structures in the condensed graph, and overlook critical issues in effectiveness and generalization ability. In this paper, we advocate a new Structure-Free Graph Condensation paradigm, named SFGC, to distill a large-scale graph into a small-scale graph node set without explicit graph structures, i.e., graph-free data. Our idea is to implicitly encode topology structure information into the node attributes in the synthesized graph-free data, whose topology is reduced to an identity matrix. Specifically, SFGC contains two collaborative components: (1) a training trajectory meta-matching scheme for effectively synthesizing small-scale graph-free data; (2) a graph neural feature score metric for dynamically evaluating the quality of the condensed data. Through training trajectory meta-matching, SFGC aligns the long-term GNN learning behaviors between the large-scale graph and the condensed small-scale graph-free data, ensuring comprehensive and compact transfer of informative knowledge to the graph-free data. Afterward, the underlying condensed graph-free data would be dynamically evaluated with the graph neural feature score, which is a closed-form metric for ensuring the excellent expressiveness of the condensed graph-free data. Extensive experiments verify the superiority of SFGC across different condensation ratios.

Xin Zheng, Miao Zhang, Chunyang Chen, Quoc Viet Hung Nguyen, Xingquan Zhu, Shirui Pan• 2023

Related benchmarks

TaskDatasetResultRank
Node ClassificationCora
Accuracy83.58
1215
Node ClassificationCiteseer
Accuracy73.14
393
Node ClassificationwikiCS
Accuracy80.08
317
Node ClassificationarXiv
Accuracy70.48
219
Node ClassificationREDDIT
Accuracy89.9
192
Node Classificationamazon-ratings
Accuracy44.02
173
Node-level classificationFlickr
Accuracy47.1
58
Node ClassificationInstagram
Accuracy65.12
34
Node ClassificationChildren
Accuracy46.83
19
Node ClassificationPokec-n-G r: 0.05 (test)
Accuracy70.45
9
Showing 10 of 15 rows

Other info

Follow for update