Yes, that is a good point. The scaling in HMC is actually much more flexible - it could be (the diagonal of) a covariance matrix or a precision matrix. We should have a detail docs to explain the differences.
Yes, that is a good point. The scaling in HMC is actually much more flexible - it could be (the diagonal of) a covariance matrix or a precision matrix. We should have a detail docs to explain the differences.