I read the paper “Unsupervised real-time anomaly detection for streaming data” carefully.
As this post " Anomaly Likelihood is always over 0.5" says, according to the mathematical formula, the anomaly likelihood is generally 0.5 under normal conditions.
But this picture in the paper shows that under normal circumstances, the anomaly likelihood is 0.0 .
I don’t know whether there is something special treatment in the program.
It really puzzled me.