Questions about the SMD dataset

[mhkim9714]

Hello. I am a fellow researcher working on interpretable time series anomaly detection. I have some questions about your work.

As I am working with the SMD dataset and the given interpretation label to measure the anomaly interpretation performance, I got some doubts about the integrity of the dataset. First of all, the start and the end timestamp in the interpretation label does not correctly match to the test label. Also, there are some missing or extra interpretations in the dataset. How did you deal with the inconsistencies when conducting the experiment?

I would appreciate it if you could clarify these points for me. Again thank you for the nice work.