Multi-class Classification on CIFAR-100 (test) (Prediction Set Size, Runtime)
8.5Average Prediction Set SizeTopK
Evaluation results
| Method | Links | ||
|---|---|---|---|
| TopKBackbone=Res-net 18, Target coverage=90%2025.05 | 8.5 | 2 | |
| APSBackbone=Res-net 18, Target coverage=90%2025.05 | 6.8 | 2 | |
| RAPSBackbone=Res-net 18, Target coverage=90%2025.05 | 6.5 | 2 | |
| PCS DropoutBackbone=Res-net 18, Target coverage=90%, Number of bootstraps (B)=10002025.05 | 4.4 | 4 | |
| PCS NoiseBackbone=Res-net 18, Target coverage=90%, Number of bootstraps (B)=10002025.05 | 4.2 | 3 | |
| PCS OriginalBackbone=Res-net 18, Target coverage=90%, Number of bootstraps (B)=1002025.05 | 3.7 | 350 |