Multi-class classification on PTB-XL (test)
92.8Macro AUCJEPA
Evaluation results
| Method | Links | ||
|---|---|---|---|
| JEPAPretrain dataset=10 datasets, Pretrain Records=1 M, Downstream trials=102026.07 | 92.8 | — | |
| JEPAPretrain dataset=MIMIC-IV-ECG, Pretrain Records=800 K, Downstream trials=102026.07 | 92 | — | |
| ER-JEPAPretrain dataset=Shao+CODE15, Pretrain Records=180 K, Downstream trials=102026.07 | 91.6 | — | |
| ER-JEPAPretrain dataset=Shao+CODE15, Pretrain Records=180 K2026.07 | 91.1 | — | |
| ECG-JEPAPretrain dataset=Shao+CODE15, Pretrain Records=180 K2026.07 | 90.3 | — | |
| KEDPretrain dataset=MIMIC-IV-ECG, Pretrain Records=800 K2026.07 | 88.8 | — | |
| 1D-JEPAPretrain dataset=Shao+CODE15, Pretrain Records=180 K, Number of pretrained encoders=52026.07 | 88.8 | — | |
| ST-MEMPretrain dataset=Shao+CODE15, Pretrain Records=180 K2026.07 | 87.9 | — | |
| ECG-FMPretrain dataset=UHN [10], Pretrain Records=622 K2026.07 | 85.6 | — | |
| SimCLRPretrain dataset=Shao+CODE15, Pretrain Records=180 K2026.07 | 83 | — | |
| MTAEPretrain dataset=Shao+CODE15, Pretrain Records=180 K2026.07 | 80.7 | — | |
| MLAEPretrain Records=180 K2026.07 | 77.9 | — | |
| MoCo v3Pretrain dataset=Shao+CODE15, Pretrain Records=180 K2026.07 | 73.9 | — | |
| ECG-FMEvaluation Protocol=Linear evaluation2024.10 | — | 85.6 | |
| ECG-JEPAmbEvaluation Protocol=Linear evaluation, Masking Strategy=Multi-block masking2024.10 | — | 90.3 | |
| ECG-JEPArbEvaluation Protocol=Linear evaluation, Masking Strategy=Random masking2024.10 | — | 89.7 | |
| KEDEvaluation Protocol=Linear evaluation2024.10 | — | 88.8 | |
| MLAEEvaluation Protocol=Linear evaluation2024.10 | — | 77.9 | |
| MoCo v3Evaluation Protocol=Linear evaluation2024.10 | — | 73.9 | |
| MTAEEvaluation Protocol=Linear evaluation2024.10 | — | 80.7 | |
| SimCLREvaluation Protocol=Linear evaluation2024.10 | — | 83 | |
| ST-MEMEvaluation Protocol=Linear evaluation2024.10 | — | 87.9 |