forestsking / memae Goto Github PK

Why did you create head and tail? Is it to make the length of prediction as equal to the row length of data? If it so, then isn't adding zeros to prediction making a false prediction that the head and tail index is a normal data?
Also, I have a question about nstep.
If nstep>1, isn't there a mismatch between the true_outlier and predicted_outlier because the gap between src and predict is calculated based on 64 data, not 1 data?