search

liznerski / fcdd

Repository for the Explainable Deep One-Class Classification paper
MIT License
224 stars 61 forks source link

Why the calculation of anomaly_scores of test data need to use its groundtruth label? #26

Closed szubing closed 2 years ago

szubing commented 2 years ago

In the trainner.test(), it seems the calculation of anomaly_scores of test data need to use its groundtruth label? It may be unresonable?

liznerski commented 2 years ago

Hey. As you can see here, the loss function ignores labels as long as the network is in eval mode, and, as you can see here, for testing the network indeed is in eval mode. So, no worries ;)

szubing commented 2 years ago

Hey. As you can see here, the loss function ignores labels as long as the network is in eval mode, and, as you can see here, for testing the network indeed is in eval mode. So, no worries ;)

My problems is solved. Thank you very much!