Table 3 AUC values (mean ± std) of different numbers of anomalous graphs in training set.
From: Deep graph level anomaly detection with contrastive learning
Number | BZR | COX2 | IMDB | NCI1 | MMP | HSE |
|---|---|---|---|---|---|---|
10 | 0.619 ± 0.049 | 0.583 ± 0.061 | 0.657 ± 0.023 | 0.683 ± 0.011 | 0.679 ± 0.042 | 0.590 ± 0.264 |
50 | 0.619 ± 0.049 | 0.584 ± 0.061 | 0.656 ± 0.023 | 0.683 ± 0.011 | 0.676 ± 0.068 | 0.590 ± 0.160 |
100 | – | – | 0.657 ± 0.023 | 0.683 ± 0.011 | 0.678 ± 0.102 | 0.586 ± 0.201 |
200 | – | – | 0.656 ± 0.023 | 0.683 ± 0.011 | 0.679 ± 0.076 | 0.587 ± 0.095 |
500 | – | – | – | 0.683 ± 0.011 | 0.678 ± 0.081 | 0.589 ± 0.105 |
No | 0.715 ± 0.067 | 0.615 ± 0.044 | 0.656 ± 0.023 | 0.683 ± 0.011 | 0.696 ± 0.042 | 0.618 ± 0.110 |
