fengduqianhe
commented
6 days ago Hi wakeupll， thanks for your question, because GGAD studies semi-supervised setting with only normal node labeled, there is no validation set here. After the anomalies are generated it becomes a classification problem, the final values can be obtained by observing the minimum cls loss function and end epoch in case of overfitting.
Dear Authors,
In Table2, you provide AUC and AP. How to get the final values? 1.End epoch? 2.Cross best evaluate?
Kind regard!