Semi-supervised classification on ADNI CN vs. MCI (10% train, 90% test)
81.5AccuracyARMA
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ARMALoss Function=Mod[29], λcon=0.012026.05 | 81.5 | 85.4 | 80.8 | 82.9 | |
| ARMALoss Function=Cut[48], λcon=0.012026.05 | 81.5 | 87.1 | 78.6 | 82.5 | |
| ARMA-C3Loss Function=–, λcon=0.00012026.05 | 79.7 | 81.9 | 82.2 | 81.7 | |
| GATLoss Function=Cut, λcon=0.92026.05 | 79.6 | 86.1 | 75.9 | 80.4 | |
| JacobiConvLoss Function=–, λcon=–2026.05 | 78.7 | 81.4 | 80.5 | 80.7 | |
| ARMALoss Function=MERIT + Cut, λcon=0.012026.05 | 78.7 | 81.2 | 80.7 | 80.8 | |
| MLPClassifierLoss Function=–, λcon=–2026.05 | 78.3 | 81 | 80.2 | 80.4 | |
| Random ForestLoss Function=–, λcon=–2026.05 | 78.3 | 80 | 81.8 | 80.6 | |
| APPNPLoss Function=–, λcon=–2026.05 | 78.2 | 80.7 | 80.8 | 80.4 | |
| ARMALoss Function=DGI + Cut, λcon=0.092026.05 | 78.2 | 83.6 | 76.2 | 79.6 | |
| GATLoss Function=Mod, λcon=–2026.05 | 78.1 | 81.4 | 80.5 | 80.1 | |
| GATLoss Function=DGI + Cut, λcon=0.92026.05 | 77.4 | 81.6 | 77.4 | 79.2 | |
| SVCLoss Function=–, λcon=–2026.05 | 77.2 | 79.2 | 80.7 | 79.7 | |
| ARMALoss Function=DGI[27] + Mod, λcon=0.012026.05 | 77.2 | 78.9 | 81.5 | 79.9 | |
| GATLoss Function=MERIT + Cut, λcon=0.00012026.05 | 77.1 | 79.4 | 80.1 | 79.5 | |
| GATLoss Function=MERIT[11] + Mod, λcon=0.00012026.05 | 76.9 | 79.3 | 79.7 | 79.2 | |
| ARMALoss Function=Baseline, λcon=–2026.05 | 75.8 | 78.5 | 77.9 | 78.1 | |
| GATLoss Function=DGI + Mod, λcon=22026.05 | 75.5 | 78.7 | 77.6 | 77.8 | |
| XGBoostLoss Function=–, λcon=–2026.05 | 73.9 | 75.6 | 79 | 77 | |
| GATLoss Function=Baseline, λcon=–2026.05 | 69 | 70.5 | 75.7 | 72.9 |