Sleep Stage Classification on Sleep-EDF
94.4Accuracyk-NN
Evaluation Results
| Method | Links | |
|---|---|---|
| k-NNInput channel=Fpz-Cz, Feature type=hand-crafted, Subjects=20 SC, Independent testing=yes, In-bed data only=no2018.05 | 94.4 | |
| Decision treesInput channel=Fpz-Cz + Pz-Oz, Feature type=hand-crafted, Subjects=20 SC, Independent testing=no, In-bed data only=no2018.05 | 93.1 | |
| Ensemble SVMInput channel=Pz-Oz, Feature type=hand-crafted, Subjects=10 SC + 10 ST, Independent testing=yes, In-bed data only=no2018.05 | 91.1 | |
| LITEAvg Params=16.84K, Avg GFlops=0.27G2025.08 | 85.3 | |
| PRISMAvg Params=13.59K, Avg GFlops=0.04G2025.08 | 85.02 | |
| Deep CNNInput channel=Fpz-Cz + hor. EOG, Feature type=learned, Subjects=20 SC, Independent testing=no, In-bed data only=yes2018.05 | 84 | |
| MambaAvg Params=235.81K, Avg GFlops=5.11G2025.08 | 83.75 | |
| TSLANetAvg Params=762.32K, Avg GFlops=2.12G2025.08 | 83.67 | |
| Multitask 1-max CNNInput channel=Fpz-Cz + hor. EOG, Feature type=learned, Subjects=20 SC, Independent testing=yes, In-bed data only=yes2018.05 | 82.3 | |
| Deep CNN + RNNInput channel=Fpz-Cz, Feature type=learned, Subjects=20 SC, Independent testing=yes, In-bed data only=no2018.05 | 82 | |
| Multitask 1-max CNNInput channel=Fpz-Cz, Feature type=learned, Subjects=20 SC, Independent testing=yes, In-bed data only=yes2018.05 | 81.9 | |
| MiniROCKETAvg Params=72.55K, Avg GFlops=0.55G2025.08 | 80.55 | |
| k-NNInput channel=Fpz-Cz + Pz-Oz, Feature type=hand-crafted, Subjects=20 SC, Independent testing=no, In-bed data only=no2018.05 | 80 | |
| 1-max CNNInput channel=Fpz-Cz, Feature type=learned, Subjects=20 SC, Independent testing=yes, In-bed data only=yes2018.05 | 79.8 | |
| Attentional RNNInput channel=Fpz-Cz, Feature type=learned, Subjects=20 SC, Independent testing=yes, In-bed data only=yes2018.05 | 79.1 | |
| Deep auto-encoderInput channel=Fpz-Cz, Feature type=hand-crafted, Subjects=20 SC, Independent testing=yes, In-bed data only=yes2018.05 | 78.9 | |
| Ensemble SVMInput channel=Fpz-Cz + Pz-Oz, Feature type=hand-crafted, Subjects=20 SC + 22 ST, Independent testing=no, In-bed data only=yes2018.05 | 78.9 | |
| PatchTSTAvg Params=485.79K, Avg GFlops=1.63G2025.08 | 77.66 | |
| ResNetInput channel=Fpz-Cz + hor. EOG, Feature type=learned, Subjects=20 SC, Independent testing=yes, In-bed data only=yes2018.05 | 76.8 | |
| Decision treesInput channel=Fpz-Cz, Feature type=hand-crafted, Subjects=10 ST, Independent testing=yes, In-bed data only=no2018.05 | 75 | |
| Deep CNNInput channel=Fpz-Cz, Feature type=learned, Subjects=20 SC, Independent testing=yes, In-bed data only=yes2018.05 | 74.8 | |
| GMMInput channel=Pz-Oz, Feature type=hand-crafted, Subjects=19 SC, Independent testing=no, In-bed data only=no2018.05 | 73.2 | |
| TimesNetAvg Params=56.69M, Avg GFlops=4.07T2025.08 | 60.18 | |
| LightTSAvg Params=1.85M, Avg GFlops=0.07G2025.08 | 57.6 | |
| FiLMAvg Params=12.59M, Avg GFlops=0.58G2025.08 | 49.66 | |
| iTransformerAvg Params=491.73K, Avg GFlops=0.05G2025.08 | 39.21 | |
| DLinearAvg Params=3.64M, Avg GFlops=0.12G2025.08 | 23.92 |