Skip to main content

Table 2 Performance of GAT2 with different neural network structures (mean \(\pm\) std)

From: GAT-LI: a graph attention network based learning and interpreting method for functional brain network classification

Number of layers Number of multi-head for each layer Accuracy Sensitivity Specificity F1 AUC MCC
1 5 0.6696 ± 0.0332 0.6416 ± 0.0440 0.6962 ± 0.0307 0.6541 ± 0.0378 0.7251 ± 0.0388 0.3385 ± 0.0668
3 5, 5, 3 0.6415 ± 0.0422 0.6435 ± 0.1286 0.6396 ± 0.0629 0.6312 ± 0.0629 0.7145 ± 0.0480 0.2923 ± 0.0796
2 5, 5 0.6676 ± 0.0409 0.6812 ± 0.0706 0.6547 ± 0.0935 0.6660 ± 0.0404 0.7261 ± 0.0384 0.3390 ± 0.0787
2 3, 3 0.6599 ± 0.0371 0.6753 ± 0.0413 0.6453 ± 0.0642 0.6597 ± 0.0337 0.7178 ± 0.0529 0.3214 ± 0.0731
2 5, 3 0.6802 ± 0.0269 0.7406 ± 0.0408 0.6226 ± 0.0534 0.6931 ± 0.0248 0.7358 ± 0.0373 0.3426 ± 0.0628
  1. The bold means it is the best result for each metric (column of the table)