Open cpaulik opened 9 years ago
This is "slightly" related to issue #12 indeed. But in issue #12, I use a pre-computed weight (e.g. an uncertainty estimate), and was not aiming at an on-the-fly decision (e.g. your decision tree).
To address your need, we must somehow encapsulate at a higher level (you do not want a weighted average). I'll have to look deep down in the kdtree resampling code.
@TomLav @cpaulik Is this issue still actual ?
I do not have a stressing need for it at the moment but it would be a nice feature to have in general.
@cpaulik ok, thanks for the feedback. PR welcome :smiley:
For some use cases the radius is not enough to do a proper resampling.
We are having this problem with categorical data which can not be averaged in a meaningful way but needs e.g. a decision tree to determine the correct aggregate.
This is probably slightly related to issue #12.