evidential DL for anomaly detection in time series

[Alessio-Siciliano]

Hello, I'm writing my thesis about anomaly detection in time series and I would like to implement your algorithm to recognize anomalies.
My idea is to calculate the train (only ID samples) entropy and then the test (ID+OOD samples) one as you did in your paper.
Since I could have time series with some features integer and not float, this can be a problem because the assumption of the method is the normality of the distribution right?

In your opinion, can be this method useful? Do you have some ideas or suggestions?

Thanks in advance!