Deep Clustering on CIFAR-10
0.863SCDCAM
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DCAMselection_criterion=Per-method best SC, RRL_constraint=within 10% of pretrained AE loss, architecture=Best across architectures2026.01 | 0.863 | -19.5 | |
| DCAMselection_criterion=Best RRL while SC is within 10% of the best SC of the method2026.01 | 0.809 | -20.4 | |
| DCECselection_criterion=Per-method best SC, RRL_constraint=within 10% of pretrained AE loss, architecture=Best across architectures2026.01 | 0.787 | 9.6 | |
| DCECselection_criterion=Best RRL while SC is within 10% of the best SC of the method2026.01 | 0.786 | -0.9 | |
| DEKMselection_criterion=Best RRL while SC is within 10% of the best SC of the method2026.01 | 0.622 | 180 | |
| NNMselection_criterion=Per-method best SC, RRL_constraint=within 10% of pretrained AE loss, architecture=Best across architectures2026.01 | 0.587 | — | |
| NNMselection_criterion=Best RRL while SC is within 10% of the best SC of the method2026.01 | 0.587 | — | |
| SCANselection_criterion=Per-method best SC, RRL_constraint=within 10% of pretrained AE loss, architecture=Best across architectures2026.01 | 0.541 | — | |
| EDCselection_criterion=Best RRL while SC is within 10% of the best SC of the method2026.01 | 0.541 | 74.3 | |
| SCANselection_criterion=Best RRL while SC is within 10% of the best SC of the method2026.01 | 0.541 | — | |
| EDCselection_criterion=Per-method best SC, RRL_constraint=within 10% of pretrained AE loss, architecture=Best across architectures2026.01 | 0.511 | 10 | |
| Agglo.selection_criterion=Per-method best SC, RRL_constraint=within 10% of pretrained AE loss, architecture=Best across architectures2026.01 | 0.372 | — | |
| Agglo.selection_criterion=Best RRL while SC is within 10% of the best SC of the method2026.01 | 0.372 | — | |
| ClAMselection_criterion=Per-method best SC, RRL_constraint=within 10% of pretrained AE loss, architecture=Best across architectures2026.01 | 0.208 | — | |
| ClAMselection_criterion=Best RRL while SC is within 10% of the best SC of the method2026.01 | 0.208 | — | |
| DEKMselection_criterion=Per-method best SC, RRL_constraint=within 10% of pretrained AE loss, architecture=Best across architectures2026.01 | 0.116 | 8.6 | |
| k-meansselection_criterion=Per-method best SC, RRL_constraint=within 10% of pretrained AE loss, architecture=Best across architectures2026.01 | 0.084 | — | |
| k-meansselection_criterion=Best RRL while SC is within 10% of the best SC of the method2026.01 | 0.084 | — |