Opened this issue 2 years ago · 0 comments
I am referring to this line. I believe the median value of the features should be computed on a per-era basis to avoid introducing biases.