Convolutional Neural Network-Based Discriminator for Outlier Detection

Table 4

The results of the 2-seed discriminator for MNIST with data augmentation.

Noise ratio	10 (%)	50 (%)	90 (%)
Recall	75.68	86.17	96.64
F1 measure	44.93	87.68	98.06
False negative	0.12	1.1	8.9
False positive	70.9	21.3	2.88