Table 3:

Comparison of model performance on age classification task

	Age classification with rs fMRI
Category	Method	ACC	SEN	SPE	AUC
Graph based models	TAG	0.7649 ± 0.0218	0.6673 ± 0.0328	0.8719 ± 0.0250	0.8441 ± 0.0205
	SAGE	0.7952 ± 0.0211	0.7176 ± 0.0334	0.8699 ± 0.0238	0.8736 ± 0.0186
	CHEB	0.8051 ± 0.0212	0.7494 ± 0.0329	0.8636 ± 0.0260	0.8645 ± 0.0193
	GCN	0.8104 ± 0.0195	0.7990 ± 0.0288	0.8214 ± 0.0273	0.8622 ± 0.0191
	GAT	0.8306 ± 0.0190	0.8225 ± 0.0281	0.8386 ± 0.0264	0.8883 ± 0.0175
Hypergraph based models	HGNN	0.8692 ± 0.0170	0.8972 ± 0.0209	0.8417 ± 0.0257	0.9219 ± 0.0141
	HGSVM-L	0.8735 ± 0.0167	0.8583 ± 0.0251	0.8893 ± 0.0233	0.9450 ± 0.0112
	wHGNN	0.8743 ± 0.0175	0.8717 ± 0.0241	0.8770 ± 0.0247	0.9290 ± 0.0141
	HGSVM-R	0.8826 ± 0.0168	0.8805 ± 0.0251	0.8849 ± 0.0231	0.9583 ± 0.0089
	dw HGCN(Ours)	0.9236 ± 0.0133	0.9460 ± 0.0142	0.8877 ± 0.0255	0.9770 ± 0.0058

ACC, SEN, SPE, and AUC are the abbreviations of accuracy, sensitivity, specificity, and area under the curve, respectively. We display the result for mean value ± standard derivation from 1000 bootstraps. The results with the best performance are highlighted in boldface. We use the same abbreviations and bootstrapping strategy to present the results in the following tables.