notesum.ai

Published at November 26

Contrastive Graph Condensation: Advancing Data Versatility through Self-Supervised Learning

cs.LG

Released Date: November 26, 2024

Authors: Xinyi Gao¹, Yayong Li¹, Tong Chen¹, Guanhua Ye², Wentao Zhang³, Hongzhi Yin¹

Aff.: ¹The University of Queensland, Brisbane, Australia; ²Beijing University of Posts and Telecommunications, Beijing, China; ³Peking University, Beijing, China

Arxiv: http://arxiv.org/abs/2411.17063v1

Dataset	$r$	$\{{\bf A},{\bf X},{\bf Y}\}$								$\{{\bf A},{\bf X}\}$	Whole Dataset
Dataset	$r$	Kcenter	Herding	Coarsening	GCond	GCDM	SimGC	GCSR	GDEM	Ours	Whole Dataset
Cora	1.30%	26.4±3.2	28.0±2.6	34.0±2.0	35.9±2.1	34.5±2.3	34.5±1.3	35.4±1.0	34.4±1.0	48.6±1.2	38.6±2.0
	2.60%	31.0±4.5	35.3±2.8	38.5±2.2	35.2±2.8	34.4±2.4	34.0±1.0	36.1±2.0	35.9±1.2	48.8±1.9
	5.20%	36.0±4.1	44.1±1.1	37.3±0.7	34.2±1.9	33.1±2.3	35.1±1.4	37.1±0.7	35.8±1.1	50.1±1.2
Citeseer	0.90%	17.2±2.7	24.7±3.2	30.1±2.5	31.3±1.6	33.2±1.8	34.4±1.2	33.5±0.8	34.1±0.9	40.3±1.5	32.4±0.8
	1.80%	25.2±2.8	30.6±4.8	29.6±2.0	31.5±1.7	32.9±1.5	33.4±1.6	33.5±2.3	34.0±1.5	40.4±1.7
	3.60%	23.5±4.6	31.5±2.9	29.3±2.0	32.9±0.6	32.6±0.8	35.3±1.1	34.3±0.9	34.5±1.1	39.8±1.6
Ogbn-arxiv	0.05%	34.1±0.3	33.5±0.9	23.9±2.0	32.7±0.7	33.1±0.2	32.7±0.3	34.0±0.4	33.6±0.6	35.2±0.7	35.1±1.1
	0.25%	34.5±0.7	36.3±0.7	33.9±0.0	34.3±0.1	35.2±0.5	33.0±0.5	34.1±0.9	34.9±0.5	37.2±0.3
	0.50%	35.3±1.2	37.4±0.4	35.0±0.2	35.9±0.3	35.5±0.8	33.1±0.2	33.5±0.6	34.8±0.8	36.4±0.5
Reddit	0.05%	59.6±1.5	59.2±0.4	43.2±2.0	59.9±1.3	58.1±1.2	59.3±1.4	58.7±0.3	59.7±1.2	67.6±0.5	63.2±0.8
	0.10%	60.5±0.4	60.9±1.0	55.7±0.9	59.0±2.0	58.1±1.3	60.0±1.5	59.7±1.7	59.9±1.4	70.7±0.9
	0.20%	60.4±1.5	61.5±1.2	60.1±1.2	62.4±1.6	61.1±1.0	62.7±0.2	61.3±1.1	62.3±1.6	70.1±0.5