Research Article
Convolutional Neural Network-Based Discriminator for Outlier Detection
Table 3
The results of the 5-seed discriminator for MNIST.
| Noise ratio | 10 (%) | 50 (%) | 90 (%) |
| Recall | 68.42 | 80.25 | 95.6 | F1 measure | 38.44 | 83.48 | 97.54 | False negative | 0.19 | 1.04 | 12.9 | False positive | 76.1 | 28.1 | 3.7 |
|
|