Table 15 Model performance with different \(\uplambda\) parameters.
From: An incremental adversarial training method enables timeliness and rapid new knowledge acquisition
|  | \(\uplambda\) = 1e−5 | \(\uplambda\) = 0.01 | \(\uplambda\) = 0.1 | \(\uplambda\) = 0.2 | \(\uplambda\) = 0.5 |
|---|---|---|---|---|---|
Accuracy | 0.9933 | 0.9893 | 0.9880 | 0.9920 | 0.9867 |
Precision | 0.9931 | 0.9890 | 0.9878 | 0.9917 | 0.9862 |
Recall | 0.9935 | 0.9894 | 0.9880 | 0.9923 | 0.9872 |
F1-score | 0.9933 | 0.9892 | 0.9878 | 0.9919 | 0.9865 |
Robust-accuracy | 0.9533 | 0.9480 | 0.9373 | 0.9440 | 0.9387 |
