Automated choice of smoothing parameter TODO:

[eldad-a]

• Functions for auto-smoothing parameter choice need to be reviewed (make sure they follow the definitions in the original publications).
• Consider keeping only the LOG variants, dropping the non-log ones (when is it preferable to use the non-log version?) -> keep LOG for backwards compatibility; output warning if non-None
• Consider removing the multi-dimensional version (what is the Reference for the current multi-dimensional implementation?). Can output warning when non-single dimension data is provided, or revert to vectorized implementation.