Request for evaluation scripts for comparison with other anomaly detection systems

[tallyAB]

The paper mentions a comparison with other unsupervised anomaly detection systems in section 3.4. Specifically, Airtag is compared with LSTM-based, Transformer-based, and Autoencoder-based anomaly detection methods described in the deep-loglizer paper. This requires an abstraction of the SDatasets in order to run them on these systems. Could you please provide us with the abstracted version of these datasets that are compatible with these systems?

Additionally, the deep-loglizer repository contains demo files for running each of the systems here. If any of these were customized for running the SDatasets, it would be very helpful if you could also provide us with the customized version of those demo files.

[dhl123]

We followed their repo and tried to get their best performance, other codes should be the same. I did not recall there is a specific abstraction but maybe I remember it wrong. If there is abstraction, probably they are tokenized logs or the abstracted version from ATLAS.