Implement this algorithm to multivariate anomaly detection

[Alessio-Siciliano]

Hello, I am writing a thesis about anomaly detection in multivariate time series and I think that this method can be very interesting.
I would ask you if you have some suggestions to adapt this method for anomaly detection.
In past weeks, I worked on the Deep Evidential Regression for univariate time series and it detects anomalies calculating the entropy for each point. In your method, what can be a possible way? Because if I have read well the paper there is no lambda parameter to control the uncertainty.

Thanks