Classification Efficiency (Conformal Prediction) on CIFAR10 ResNet20 (test)
2.6948Efficiency RateSCP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SCPTransductive size (n)=9, Alpha=0.1, Noise level (1-epsilon)=1, Backbone=ResNet202025.09 | 2.6948 | 0.4172 | 88 | |
| APSTransductive size (n)=6, Alpha=0.1, Noise level (1-epsilon)=1, Backbone=ResNet202025.09 | 0.8913 | 0.3333 | 98 | |
| APSTransductive size (n)=9, Alpha=0.1, Noise level (1-epsilon)=1, Backbone=ResNet202025.09 | 0.8312 | 0.5094 | 92 | |
| TSCPTransductive size (n)=9, Alpha=0.1, Noise level (1-epsilon)=1, Backbone=ResNet202025.09 | 0.6068 | 0.4542 | 92 | |
| SCPTransductive size (n)=6, Alpha=0.1, Noise level (1-epsilon)=1, Backbone=ResNet202025.09 | 0.5546 | 0.4308 | 84 | |
| SCPTransductive size (n)=3, Alpha=0.1, Noise level (1-epsilon)=1, Backbone=ResNet202025.09 | 0.4368 | 0 | 88 | |
| TSCPTransductive size (n)=6, Alpha=0.1, Noise level (1-epsilon)=1, Backbone=ResNet202025.09 | 0.4316 | 0.3333 | 88 | |
| TSCPTransductive size (n)=3, Alpha=0.1, Noise level (1-epsilon)=1, Backbone=ResNet202025.09 | 0.2932 | 0 | 94 | |
| APSTransductive size (n)=3, Alpha=0.1, Noise level (1-epsilon)=1, Backbone=ResNet202025.09 | 0.288 | 0 | 96 |