datamllab / tods

TODS: An Automated Time-series Outlier Detection System
http://tods-doc.github.io
Apache License 2.0
1.47k stars 193 forks source link

How to distinguish between train/test data? #107

Open ainsuotain opened 1 year ago

ainsuotain commented 1 year ago

Hello authors,

Thanks for great code, But I still don't understand how to split the train and test dataset for training, validation and, testing In the section 4.1 , 'We conduct benchmark experiments in unsupervised setting. Each of the algorithm is trained and tested on the same dataset'

Do you mean exact same data(normal data with anomalies)? If so, please explain in more detail For example, how do you perform hyper parameter optimization on models that need validation? Autoencoder, for example, also can train the anomalies then, after few epoch model cannot discriminate the outliers

Again thank you for your great works. It helps me a lot