Research Article
Convolutional Neural Network-Based Discriminator for Outlier Detection
Table 14
Average test accuracy on CIFAR-100 over the last ten epochs.
| | Method | Flipping rate | | Symmetry-20% | Symmetry-50% | Pair-45% |
| | Standard | 47.55 (±0.47) | 25.21 (±0.64) | 31.99 (±0.64) | | Bootstrap | 47.00 (±0.54) | 21.98 (±6.36) | 32.07 (±0.30) | | S-model | 41.51 (±0.60) | 18.93 (±0.39) | 21.79 (±0.86) | | F-correction | 61.87 (±0.21) | 41.04 (±0.07) | 1.60 (±0.04) | | Decoupling | 44.52 (±0.04) | 25.80 (±0.04) | 26.05 (±0.03) | | MentorNet | 52.13 (±0.40) | 39.00 (±1.00) | 31.60 (±0.51) | | Coteaching | 54.23 (±0.08) | 41.37 (±0.08) | 34.81 (±0.07) |
| | EBF | 59.90 (±0.66) | 48.51 (±0.61) | 32.65 (±0.60) | | Discriminator (10S) | 24.45 (±0.83) | 20.93 (±0.62) | 21.42 (±0.88) | | Discriminator (30S) | 44.36 (±0.47) | 37.70 (±0.67) | 39.26 (±0.81) | | Discriminator (50S) | 49.31 (±0.67) | 41.79 (±1.11) | 44.62 (±0.54) |
|
|
